Tatem et al's article claims that sometime around 2156 women will start beating men in the 100 meters. Their figure that's supposed to show this is reproduced below:
(blus is men, red is women)
In fact, what they've managed to produce is a pretty good example of what goes wrong when you just blindly curve-fit without thinking about what you're doing. The model that Tatem et all have fit is a linear model. They defend this by arguing that none of the other models they tried produced a better r^2, but that's not really that compelling an argument. The linear model is quite clearly wrong because it produces a ludicrous result: in 2649, the Women's record for the 100m will be zero seconds!
At this point, it would be reasonable for you to object that it's ridiculous to extrapolate out the 75 years or so of data that we have for the women's 100m for another 650 years, but what makes you think it's appropriate to extrapolate it for 150 years? It's clear that although the curve may appear to be linear now it's got to flatten out somewhere. The only question is where that somewhere is. However, you can't just wish the problem away by assuming that the data is linear over the region of interest.
What's particularly annoying is that this isn't the first time that this error has been made on this topic--indeed, in this same journal. Whipp and Ward's 1992 Nature article predicted that women's marathon times would intersect men's by 1998. In reality, however, the women's marathon improvement flattened off severely and the current men's marathon WR is 2:04:55 (held by Paul Tergat) and the women's best is 2:15:25 (held by Paula Radcliffe). 2:15:25 is an outstanding time but pretty far away from 2:05. For comparison, there have been 714 men's performances under 2:10 in the last 20 years.
In fact, there's a gap of roughly 10% across the entire distance range, as shown in the following table:
|Distance (m)||Men's WR||Women's WR||Gap (%)|
The gap has actually increased from the 80s to the 1990s (if you exclude the marathon which women didn't start running in the Olympics until 1984) which suggests that it's probably not going away any time soon. It's sometimes argued (using the marathon as an example) that women are better in really long distances, but this doesn't seem to be the case. This gap also extends to other sports such as Ironman distance triathlon, with finishing times closer to 8-9 hours.
For more information, you might check out this interesting article in the July 30 Science about the persistence of the gender gap. The authors actually quote Whipp, who doesn't seem ready to admit error:
Whipp says he's still keeping an open mind on the subject of male female
competition. He told Science that he and colleagues are currently working to extend their analysis of world running records to 2003. His team has looked for a pattern of response that would suggest a physiological limit, he says, but so far has found none. There is no evidence at the beginning of the 21st century that the human athlete has reached the limit of [his or her] potential, Whipp says.
I'm keeping an open mind, too, but I don't find either Whipp's or Tatem's work at all convincing.
According to the NYT the US once again has the World's Fastest Computer. The new IBM computer has been benchmarked at 36.01 teraflops on Linpack. This is all of .4% faster than the previously World's Fastest Computer, NEC's Earth Simulator, which ran at 35.86 teraflops.
In general, I'm pretty hard pressed to care about this kind of stuff, which seems to me to fall more into the same category of US vs. USSR Olympic medal count than anything else. It's not like it's going to make my Web browser run any faster. However, in this particular case, there is something interesting: the new computer is a lot more efficient:
The new system is notable because it packs its computing power much more densely than other large-scale computing systems. BlueGene/L is one-hundredth the physical size of the Earth Simulator and consumes one twenty-eighth the power per computation, the company said.
The BlueGene/L will have wide commercial applications, first in the petroleum and biotechnology industries, Mr. Turek said.
A large-capacity version of the BlueGene/L system is scheduled to be installed early next year at the Lawrence Livermore National Laboratory in Livermore, Calif. That machine will have about 130,000 processors, compared to the 16,000-processor prototype that set the speed record.
In other words, scientific labs might be able to buy computer power more cheaply. That's really a much more interesting story than a .4% speed improvement in some benchmark.
It's pretty easy to get confused if you go to the store to buy an HRM. For instance, Road Runner Sports carries 17 separate HRMs from four different brands: Polar, Nike, Timex, and Mio Sport--and this isn't even all the monitors made by those companies. Here's the good news: (nearly) all HRMs do basically the same job. They capture your heart rate from the chest strap and put it on the watch. The main difference is in what else they can do.
To get a feel for this, let's take a quick look at some of the Polar HRM line, starting from the bottom.
* The F1 is their most basic HRM. It displays your current heart rate and your current exercise time. When you're done it will display your average heart rate.
* The A3 has all the features of the F1, plus it has alarms that help keep you in your heart rate zone. You set the limits (it will also do it automatically, but I have no idea how and don't trust this feature) and then it beeps when you're above or below the zone. It also has an actual watch feature so it knows what time it is as well as how long you've been exercising.
* The A5 is a fancier version of the A3 with more sophisticated automatic zone-setting features. I don't think this buys you much over the A3.
* The M32 is a more hard-core version of the A5. Basically, the difference is the transmitter. The original Polar transmitters were a plastic chestpiece that went on the front of your chest and an elastic strap on the back. This wasn't that uncomfortable but the M32's transmitter has a smaller plastic pod (for the radio and electronics) and then the electrodes are built into the strap so it's supposedly softer and more comfortable. Also, the transmitter is "coded" so that there's supposed to be less risk of interference between you and other HRM users. (With old-style HRMs you can pick up other people's signals sometimes).
* The S120 is sort of a step sideways. It's got most of the same features as the A3 but has fewer automatic zone-setting features. You can also program workouts into the HRM from your computer. This isn't really that necessary.
* The S150 is the same as the S120 except that it can also be used as a bike computer. Useful if you cycle.
* The S410 lets you download workout data to your PC. It appears to just record the same basic data as the other ones (time, time in HR zone, avg HR, etc). It also seems to have a lap timer so you can get per-lap data.
* The S610i seems to be the next big step up. Instead of just recording per-lap data, you can record your HR evey 15-60 seconds. This is sort of useful if you're racing and want to see your performance over the race.
* The S625X adds a running distance and pace feature using a foot-mounted transmitter pod. Timex also makes a similar unit but using a GPS instead of a pedometer. I don't think Polar has one of these, though.
I've only covered about a third of the monitors Polar makes, but this should give you an idea of what's out there. Next, I want to talk about what features you should care about.
You should at minimum look for an HRM that has programmable zone alarms. These alarms let you tell the HRM the HR range you want to maintain and it will beep when you go out of the range. This makes it easy to stay in the zone without having to constantly look at the monitor. You can get away with an HRM without this feature, but life will be easier if you have it. You want zones you can set manually because, as I said earlier, I don't trust the HRM to guesstimate your zones for you.
You probably also want a stopwatch. This makes it much easier to time your workouts, especially if you find yourself having to take breaks.
These are all the features you need for basic training. If you plan to be doing advanced cardio training (especially intervals) there are some other features that are nice to have. The first is interval timers. You can program the watch to beep every X minutes to let you know when to start an interval. It's most convenient if you have two timers so you can program one for your work interval and one for your rest interval.
If you're doing intervals it's also nice to have a "lap" memory. This lets you push a button at the start of each interval period. When you're done, you can recall each period's length and the heart rate associated with that interval. (Sometimes this is the average and sometimes it's the end of the interval. Either is pretty useful).
In the nice to have category is the ability to remember multiple workouts. If you're maintaining a training log it can be convenient to have the HRM remember what you did last week if you forgot to write it down.
You very much want to be able to replace the batteries in the chest strap. The old Polars didn't let you do that--you needed to send it back to the factory for replacement, which was inconvenient and expensive. So, check before you buy. Replacing the watch battery is easier--generally any jeweler can do it--but they often don't warrant waterpoofness. I've never had a problem, though.
If you're cycling, it can be nice to have an HRM which can also act as a bike computer (speed, etc.) You should also get one that has a cadence feature to help you develop good cycling form. If you're running, you might find it useful to have the GPS or pedometer distance measurement. I don't personally find this that useful--I usually know my own pace and if I really care I just bike the course to measure it.
Sometimes people complain about the comfort level of the plastic chest strap. I've heard that Polar's new cloth strap is more comfortable. You might check it out.
I'm not a big fan of downloading and periodic data recording. In theory you can upload a workout or race and then go back and review how you were performing in various sections. In practice, I've never found this that useful, but some racers swear by it. I suspect it's more useful if integrated with the bike computer so that you can match up your performance to the distance you've gone.
I've never had a monitor that let you upload workouts onto it. I suppose that's useful if you find the watch UI really complicated, but I've never had any big problems with that.
One annoying feature of HRMs is that they don't always work perfectly. It's pretty common to get ridiculous readings (e.g., HRs of 225 when you're jogging easy). There appear to be several causes of this:
The problem with transmission errors is simple: the old HRMs used analog signalling for the heart rate data, so various kinds of RF interference, such as power lines, seemed to screw them up. The new coded transmitters are supposed to be better. They also have the advantage that they don't pick up other people's signals so you don't accidentally see your training partner's HR. Unfortunately, in my experience transmission errors are not the most common problem, so I'm not sure how much the coded transmitters help.
It's quite common to have contact failures. The transmitter belt needs to be seated properly and the contacts need to be wet to pick up your HR. You're supposed to wet them (I use saliva) before you put them on, but if they're not wet enough then you tend to get random looking readouts. If you find that the HRM only works when you're warm and sweating, the problem is probably contact failure.
I've also seen anomalous readings--very high HRs when I'm not working hard. I've never been able to debug these properly but every HRM I've ever had does it. It doesn't happen that often and seems to go away when you're warmed up, so I suspect it has something to do with the HRM having trouble reading HR signals when your heart isn't working hard, but I just lump it into "unknown causes".
Just to get you started, here are some HRMs that would be pretty good for the beginner:
All of these watches seem to have stop watches and settable zone alarms, which I think are the key features. The Nike and Timex both seem to let you replace the batteries yourself. With the Polar you need to send it back, which is annoying. I've never tried the Timex but it looks like a good bargain.
If you want more advanced features such as distance monitoring or downloading, you'll need to do your own research (though comments on this post with more info would be welcome).
A lot of people worry about chest strap comfort. You can buy non chest strap monitors such as the Mio. These work buy having a pair of electrodes that you put your fingers on to get a reading. I don't advise using one of these because it's much more convenient to have continuous readout. Besides, the chest strap isn't noticeable once you get used to it.
Acknowledgement: Thanks to Steve Purpura for reminding me to mention using saliva to wet HRM contants.
This post covers the basic strength part of our basic workout plan.
Strength training is about your ability to do a lot of work in a short period of time. The canonical example of a strength exercise is weightlifting, but lots of sports have strength components. For instance, playing football or wrestling requires being strong in order to make people do what you want. Similarly, tennis and baseball players want to be strong in order to improve the amount of energy they can impart to the ball.
Unlike cardiovascular fitness, strength is not a monolithic property that can be trained with a single exercise. A good strength program consists of a variety of exercises, each of which exercises some of your muscles or muscle groups. A complete program will cover as many muscle groups as possible and allow you to develop evenly.
The best way to strength train is to lift weights. However, that only tells part of the story. Like aerobic training, strength training can be performed at a variety of different intensity levels, which produces different training adaptations. If you want to get very strong for short periods of time, you typically do only a few repetitions with a very heavy weight. If you want to get somewhat stronger while increasing your muscular endurance (your ability to continue to apply force over longer periods of time) you do larger numbers or repetitions with smaller weights. Our program is more balanced towards the second kind of development, which is more practical for beginners because it can be performed with less total time investment and risk of injury.
We should start by introducing some terminology. A weight lifting program consists of a number of exercises. For instance, your program might include bench press, squats, and leg extensions. Each of these is an exercise. Let's take bench press as an example.
picture from Ball State University Biomechanics Lab
To do bench press, you lie on your back on the bench with a weight bar above your head. You then lower the bar to your chest and then push it upwards until your arms are straight. This is one repetition (rep) of bench.
Exercises are organized around sets. A set is a series of reps performed together. So, if you lay down on the bench, did 5 reps and then stopped, that would be a single set. Typically, you rest between sets. The most common rest interval is between 30 seconds and 2 minutes, depending on how hard you are working and what your training goals are.
Sometimes you will combine more than one set into a superset. For instance, to do a superset of bench and then lat pulls you would do a set of bench and then go right over to the lat pull machine and do set of lat pulls. Then you would rest. This is one superset.
Unlike aerobic exercise, the most effective way to train strength is to work to failure. This means you have to get to the point where you cannot do another rep. Until you get used to it, failure can be surprising and very unpleasant. Even when you are experienced, it's never fun but it's part of lifting.
In general, we will be doing two sets of each exercise. In the first set, we will do slightly less weight and slightly more reps. In the second set, we will do a little more weight (typically about 10% more) and a fewer reps. When you are at the right weight you should be able to complete the first set and should struggle to complete the second set.
Here is a typical day of bench press used by one of my lifting partners:
Last week was the first time he had tried this amount of weight and while he successfully completed the first set, he only managed three reps in the second set. I gave him just enough assistance that he completed two more reps. When you assist someone like that it's called a forced rep. The next time we go to the gym I expect him to be able to do perhaps four reps and I'll help him through two or three more. Eventually, he'll be able to complete 8 without any assistance from me. At that point we will move up in weight to 105 and 115. There is a general principle at work here: as soon as you can easily complete the assigned set you increase the weight. It's true that this means that you will now most likely be failing again, but you just have to think of that as a sign that you're improving. Once you get into it it can be very satisfying.
For reference, the world record for the bench press, held by Scot Mendelson, is 713 pounds (965 pounds with a bench shirt but bench shirts are kind of cheating). The benchmark for being reasonably strong is being able to bench your own weight. Most men can work their way up to 225 with sufficient and at 275 you're starting to be pretty serious.
Most of the exercises in this program are designed to be done either with free weights (barbells and dumbbells) or with cable machines (a cable and a handle attached to a stack of weights). Free weights provide a better overall workout because they force you to stabilize the weight, thus working secondary muscles as well as the main muscles. Many gyms will have machines that simulate exercises with free weights.
There are two caveats in the use of free weights:
With these caveats in mind, I recommend that you use the free weights where possible. It gives you a better workout and is in general a much more satisfying experience.
There are an amazing number of different exercises with slightly different hand grip, form, tempo, etc. In some cases we suggest that you vary the exercise somewhat between the first and second set. This will give you better muscle coverage and improved strength gain. It's not required but it does help some.
If you work out in a serious gym, you'll probably see guys that work out for 3 hours at a time. They're there when you get there and they're there when you leave. It's tempting to think that more is better, but in this case it's not. Research shows that after 45-60 minutes of weightlifting serum cortisol levels start to increase and testosterone levels start to decrease, indicating that the additional training benefit is starting to decline. This has two implications. First, it means that an hour-long weight workout is plenty. Second, it means that you shouldn't dawdle, but rather try to complete your workout within an hour to get maximum training benefit.
In this section we describe the basic program, which consists of 24 sets. You should take 1 minute between sets, thus letting you do the entire program in about an hour. Each exercise has a link to instructions on how to perform it. When you see "10, 8" in the Notes that means that the first set should be 10 reps and the second 8 with increased weight as described above.
|Pull-Ups||2||To failure. Vary hand grips between sets. I would start with moderate grip and move to neutral/hammer grip|
|Dips||2||To failure--superset with upright rows|
|Hammer curl||1||10, superset with rope extensions|
|EZ bar curl||1||8, superset with bent dumbbell extension|
|Bent dumbbell extension||1||8|
|Leg extension||2||10,8 superset with leg curl|
|Hyperextensions/back crunches||2||To failure, superset with crunches|
The first 2 weeks you go to the gym you should not work to failure. Just get the feel of the weights and do weights that are comfortable. You should finish each set feeling that you could do 1-2 more reps but that would be about it. This lets you start to build a modest amount of strength and form without letting you get too sore or injured.
Over the next two weeks you should gradually increase the weights until you are failing on the last set of every exercise. Ideally, you should have a spotter with you and have them force you through the last 2-3 reps if necessary to make sure you complete the set. This maximizes results--while also somewhat maximizing suffering, which in this case is good for you.
Pull-Ups, dips, hyperextensions, and crunches all use your body weight as the source of resistance. However, while most people can do at least some hyperextensions and crunches, many people cannot do any pull-ups or dips. This is especially true of women, who tend to have less upper body strength. You will need to work into these exercises slowly.
If you can't do any pull-ups, you should start with lat pulldowns. Use the 10,8 pattern as we discussed previously. Periodically, you should try to do a few pull-ups (hammer grip is generally most comfortable). When you can do 3 pull-ups, shift to doing one set of pull-ups followed by a set of lat pulls (try for 8). Once you can do 5 or more pull-ups, you should do two sets of pull-ups as listed above.
If you can'd do any dips, you can get started by doing them on a bench instead. With your feet on the ground put your hands behind you on the bench. Then gradually lower your ass towards the floor (bending your arms) and then extend your arms to push yourself up. Do as many of these as you can. As with pull-ups, periodically try some real dips to see if you can do them. Once you can do 3-5 dips you should gradually transition to those.
Sore muscles are an inevitable result of lifting weights. It takes a while to learn the difference between injury and soreness, but trust me that there is one. Soreness is a natural result of working yourself hard and is nothing to worry about. However, it can be unpleasant. 400 mg of ibuprofen (advil) 3-4 times a day can do a lot to relieve soreness.
Once you have reached that point, keep increasing weights as each set becomes easy for you. Initially you'll make a lot of progress, which is very nice, but eventually you'll start to plateau (not make as much progress, if any.). At that point you can either shift to maintenance mode, just doing the same lifts, or go in for a more advanced program than we cover here.
Acknowledgement: Thanks to Kevin Dick who got me lifting seriously, designed the original version of this weight program, and provided many useful comments on this version.
In the previous two parts, I introduced the idea of a workout plan and divided workouts into two categories: cardio and strength. In this post, I'll be describing the basic cardio part of our workout plan.
Cardiovascular fitness is about the ability to sustain low to high moderate intensity levels for long periods of time. Most strictly cardiovascular events are also endurance events--long races. However, there are also a large number of events which combine cardiovascular fitness with strength and other skills. For instance, wrestlers have a very high level of cardiovascular fitness but must also be strong and have grappling skills.
"Long periods of time" spans nearly two orders of magnitude, everywhere from running the half mile (less than two minutes) to ultradistance running (30 hours or so for the Western States 100). We're primarily concerned with the low middle portion of that range, say 30 minutes to 2 hours. These efforts are aerobic, which means that to a great extent your performance depends on your ability to process oxygen. This makes your breathing and cardiovascular systems a critical performance element. We'll come back to this point later.
As you can see, the time range of aerobic activity varies dramatically. The intensity level of the effort varies as well. For example, the world record for the 800 meters (held by Wilson Kipketer) is 1:41.11 (3:22/mile). (Note: I'm phrasing a lot of this in terms of running, but the principles are generally applicable to any endurance event.) The world record for the marathon (held by Paul Tergat) is 2:04:55 (4:46/mile). On a more earthly level, I can run a single mile in about 5:30, but my marathon personal best is 3:11 (7:17/mile) and I can maintain 8 minute miles almost indefinitely. All of these paces qualify as aerobic efforts.
Interesting note: the 200m WR pace is slightly faster than the 100m pace. This is because a lot of the 100m is taken up by the start. In the 200m you have more time to amortize the starting cost. As you can see, you start to get tired in the 400m, so the pace is a little slower).
If you're going to race fast, you occasionally need to train fast, but the way that you train isn't to go out and run your preferred distance as fast as you can. Going fast involves a number of different physiological adaptations and you get the best training stimulus for each of those adaptations at a different intensity level. Serious endurance athletes will typically train at something around four different levels, listed here in increasing order of intensity:1
A serious endurance athlete would train at all these levels. However, somewhat surprisingly for novices, generally at least half the training is done at the easy overdistance pace because Overdistance stimulates some very important physiological adaptions, especially for general fitness. A typical training load might be something like 55% Overdistance, 15% Endurance, 25% Intensity, and 5% Speed.
If you're not racing, you can get a very large amount of fitness benefit from Overdistance training alone. This is nice because Overdistance training, though boring, is very easy, so you don't have to suffer that much. It's also the best way to start a cardiovascular training program without hurting yourself. Our initial training program will be Overdistance only, but I'll also describe some advanced variants that introduce a little Endurance and Intensity work (Speed isn't that important if you're not racing).
When you work hard, your muscles need to produce more energy. As I mentioned earlier, this energy-production process uses oxygen, so your body needs to deliver more oxygen to the muscles. In order to do this, you start to breathe faster (to get more oxygen into the blood) and your heart starts to beat faster (to pump that oxygenated blood to the muscles). Because of this, you can use your heart rate as a reliable guide to the level of effort. This is the principle behind the Heart Rate Monitor (HRM), the single most important training tool for the endurance athlete. You will need to buy one.
An HRM is basically a single-purpose, portable version of the electrocardiogram machines used in doctor's offices. Most HRMs have two pieces:
The simplest HRMs just tell you your current heart rate. More advanced ones have features like a stopwatch, alarms for various heart rate levels, memories, etc. I'll talk a little about how to buy a HRM in a future post. You can also ask your local running or cycling store.
The way you use an HRM is simple: you figure out in advance what the heart rate range you want to be in for any given workout is and then you use the HRM to adjust your effort level to stay in that zone. If your HR is too high, you need to slow down. If it's too low, you need to speed up. The only tricky part is figuring out your heart rate zones. We'll cover that next.
Everyone's natural heart rates are a little different. So, in order to find the appropriate training zones we need to figure out where your natural heart rates are. For the moment, there are two measurements we care about:
Finding your resting heart rate is easy. Take your heart rate monitor to bed. When you wake up in the morning, put on the monitor and measure your heart rate. Do this for 5-7 days in a row and take the average (mean) as your RHR.
Finding your MHR is more tricky. A lot of people like to estimate it as 220 - Age, but this formula can be wildly inaccurate and it's better to measure. Getting an exact measurement is a little tricky, but for our purposes a rough measurement is good enough. Here's a simple 1/2 mile test that you do wearing your heart rate monitor.
First, find a track in your area. Warm up by running for 10-15 minutes fairly easy. Stretch if you need to. Now you're ready for the test.
After this test, you should probably walk a lap or two to cool down.
Once you know your RHR and MHR, you can compute your training zones. This is done using what's called the Karvonnen Formula:
The way you use this formula is to decide on your effort level (expressed as a percentage) and then use it to generate your heart rate.
The effort levels for the training zones we've mentioned are (approximately):
For example: My RHR is about 50. My MHR is about 195. So, this gives us the following heart rates:
|Zone||Effort Range||HR Range|
You can find a Karvonnen formula calculator here.
In our basic workout plan, we will be doing Overdistance workouts only. Our target is to have you doing 3 workouts a week of 40-50 minutes in the Overdistance zone. Doing these workouts is really simple: run, or bike, or do the stairmaster for the required time while keeping your heart rate in the zone.
Let me say that again: keep your heart rate in the appropriate zone. If you've done any aerobic training before, these workouts are going to feel incredibly easy by comparison and it's tempting to go harder. Resist that temptation at all costs. You will get the maximum training benefit with the minimum risk of overtraining if you stay in the zone. Once you get used to working out with an HRM, you'll get the feel of Overdistance, but until then you have to trust the HRM.
If you're not used to doing this kind of training, you should not start by doing the full workout. You need to work up to it slowly. This training plan assumes that you're in it for the long haul, so there's no need to rush things. We're not going to start by going easy--this effort level is easy enough as it is--but by going short and building up the training volume. For the rest of this section, I'm going to assume that you're starting from nothing, so we'll start with very short workouts. If you find that's too easy, you can jump ahead a bit, but don't get too aggressive. It's far better to build up slow and stay uninjured than to get two weeks of good training in and then be too injured or sick to continue.
There are two ways to build up your workout volume:
The second strategy is far better. First, if you're not in good shape, a 40 minute workout will be too much. Second, it's good to get into the habit of going three days a week as soon as possible.
I recommend that you start by working out three days a week for 20 minutes each. You can then build up by 10% each week until you hit the target level. This sounds like a very slow buildup, but it's actually fairly fast while remaining safe. Inside three months you'll be up to 50 minutes a day.
|Week||Workout Time (min)|
One thing that you'll notice as you build up is that you find yourself going faster in a given heart rate zone. This is natural and shows you're getting more fit. You just need to stay in the zone.
During this build-up period, you need to be very careful to watch out for signs of incipient injury. Some aching is natural as you get used to working out more, but if something starts to really hurt, you should be concerned. One good technique for reducing the risk of injury is to do different exercises each day. For instance, you might do exercise bike one day and treadmill the next. This reduces the risk of overuse injury due to one specific exercise. This is particularly good advice if you are running, because running can be very hard on your joints and the adaptation takes a long time.
One thing you may find surprising is that I haven't specified a warmup period. It's not necessary here because Overdistance training is done more or less at warmup pace. If you find yourself being a little stiff initially, just start out slow for the first few minutes until you warm up.
One effect that many people notice is that their heart rate climbs during a workout. This is very normal initially because it takes a few minutes for your body to get warmed up. After about 10 minutes it should stabilize, but then will often slowly creep up after a half hour or so. This can either indicate that you're getting tired (normal) or that you're gettng dehydrated. Make sure to drink plenty of fluid, especially as your workouts pass the 40 minute mark. This is a special concern if you're working out indoors, where it tends to be hot and you often don't get good airflow for cooling. In any case, be careful not to let your heart rate creep out of the zone. Slow down if you need to.
Running is a very popular aerobic exercise and a great workout, but it's also a good way to get injured, especially if you've never run before. If you're going to run, you absolutely must get quality gear. I strongly advise going to a running specialty store and having them fit you for a good pair of shoes (this is also a good time to buy an HRM). A good store will watch you run and help you pick out a pair of shoes that match your stride. If you live in the South Bay, I recommend The Runners High.
Running equipment is not a place to be cheap! Good shoes make the difference between getting injured and staying healthy. Expect to spend $80-100 on your shoes, especially if you're heavy (shoes for heavy people tend to be more expensive, no doubt because they have more science in them). You should also invest in some good technical socks to avoid blisters. You also need to replace your shoes regularly, typically every 400 miles or so. I write the purchase date on my shoes and use that to know when to replace them.
Many people find that when they run their HR quickly rises out of the zone. The solution here is to run/walk. Start by running for 5 minutes and then walking for 5. Do this for your entire workout (the walking counts as part of your time!). Each week run one minute longer and walk one less until you're running the whole time. You should keep the same total workout time until you're running the entire distance, so using this plan you would do 5 weeks of 20 minute workouts. Once you get up to running all the time you can start ramping up your volume.
Our target training time is 40-50 minutes per workout, three times a week. That's a perfectly fine level and once you reach that it's perfectly reasonable to just stay there. However, if you start to feel like you want more, don't just keep ramping up the volume. It's time to start using the other training zones. I'll talk about that in a future post.
1 The taxonomy I'm using here is a kind of compressed version of that used by Sleamaker and Browning in Serious Training for Endurance Athletes. If you're interested in training seriously for endurance sports, I highly recommend this book. What I've done is compress Intervals, Up/Vertical into Intensity and done without Race/Pace pretty much entirely. If you're a racer, you'll also do Race/Pace or Tempo efforts where you try to go at pretty much your maximum sustainable pace over a given distance. This helps you practice for races and measure your fitness but doesn't have much place in a general fitness program. Another good reference (though a big dated) is The Heart Rate Monitor Book
2 Typically MHRs for running are a little higher than for cycling or other sports where you sit, but the difference is only important for serious athletes.
3 There are more complicated tests where you progressively ramp up the load for 10-20 minutes or so until you fail. These will give you a more accurate reading but are much less fun. If your doctor has given you a "stress test", this will also typically give you your MHR.
Spent a few hours at Terence's playing Def Jam Fight For NY. Basically, it's a fighting game, but the hook is that the competitors are mostly rap stars and other celebrities. In DJFFNY you've got two choices: you can either play one of the stock characters or you can create your own fighter.
The interface for creating your own fighter is actually pretty cool. It's modelled (at least loosely) on the process the police use for creating composite pictures. You start by picking the body type, then move on to head shape, eyes, eyebrows, etc. My character, Ice-E, looked scarily like I would if I were a street thug. Once you've created your character, you can enter a story line where you fight your way to the top. You win points and money by winning fights and then you can use the points to improve your skills--in a gym run by a very accurate looking Henry Rollins. Alternatively, you can just play (at least in one-on-one mode) using any of the available characters.
As usual, the controls are basically incomprehensible, at least to beginners, and your skill level seems to matter less than what character you choose. Terence clobbered me repeatedly until I played Henry Rollins against his Lil' Kim and whupped him. Definitely worth a try, if just for the entertainment of watching various characters face off. Next time, I want to have Danny Trejo fight Flavor Flav.
The other day I was reminded of the famous SGI 5.1 Memo. One of the big complaints about IRIX 5.1 was how bloated it was:
In the May report, I listed a bunch of executable sizes, and pointed out that they were unacceptable if we intended to run without serious paging problems on a 16 megabyte system. Between May and the 5.1 release, many have grown even larger. IRIX went up from 4.8 megabytes to 8.1 megabytes, and has a memory leak that causes it to grow. Within a week, my newly-booted 5.1 IRIX was larger than 13.8 megabytes -- a big chunk of a 16 megabyte system. It's wrong to require our users to reboot every week.
There are too many daemons. In a vanilla 5.1 installation with Toto, there are 37 background processes.
DSOs were supposed to reduce physical memory usage, but have had just the opposite effect, and their indirection has reduced performance.
Programs like Roger Chickering's "Bloatview" based on Wiltse Carpenter's work make some problems obvious. The news reader "xrn", starts out small, but leaks memory so badly that within a week or so it grows to 9 or 10 megabytes, along with plenty of other large programs. But what's really embarrassing is that even the kernel leaks memory that can't be recovered except by rebooting!
Showcase grew from 3.2 megabytes to 4.0 megabytes, and the master and status gizmos which are run by default occupy another 1.7 megabytes. Much of this happened simply by recompiling under 5.1 -- not because of additional code.
The window system (Xsgi + 4Dwm) is up from 3.2 MB to 3.6 MB, and the miscellaneous stuff has grown as well. As I type now, I have the default non-toto environment plus a single shell and a single text editor, jot. The total physical memory usage is 21.9 megabytes, and only because I rebooted IRIX yesterday evening to reduce the kernel size. Luckily, I'm on a 32 megabyte system without Toto, or I'd be swamped by paging.
Much of the problem seems to be due to DSOs that load whole libraries instead of individual routines. Many SGI applications link with 20 or so large DSOs, virtually guaranteeing enormous executables.
In spite of the DSOs, large chunks of Motif programs remain unshared, and duplicated in all Motif applications.
For comparison, here are some process sizes on my FreeBSD 4.9 machine, derived using top.
|rpc.statd||257M||268K||The status daemon, it's got a huge mmapped segment|
|xemacs||162M||149M||My editor (and mail reader and news reader)|
|XFree86||140M||91M||The X server1|
|xterm||4M||2M||Terminal windows (I have 10 of these open)|
Top claims that I have 400M of memory in "Active" state.
The great thing is that this machine has 1G of physical memory (total cost < $500), so this level of memory consumption is considered perfectly acceptable. To a first order, programmers have stopped worrying about program size since it's so much easier to just tell people to buy more memory. Even genuine memory leaks are a small consideration unless you have a really long-running program with a big leak. Thanks, Moore's Law!
1 Hovav Shacham points out "there's a traditional asterisk next to X server memory usage ... it's alleged to mmap the GPU device memory, which means it's using less real memory than top claims. dunno if that's actually still true." I have a 64 M graphics card, so even so, real memory consumption is pretty substantial.
Brad DeLong points out just how bad the budget deficit situation is looking:
You used to hear a lot of talk about how the purpose of large tax cuts was to force corresponding spending cuts, aka "starving the beast". So, how large do deficits have to be before we see the advocates of tax cuts try to cut spending?
Apparently, our friends at the California Air Resources Board have decided to do something about global warming:
Under the regulations, unanimously approved Friday by the California Air Resources Board, the auto industry must cut exhaust from California's cars and light trucks by 25 percent and from larger trucks and sport utility vehicles by 18 percent.
"In the short term we probably won't see much effect because global warming is a very long-term problem," said Terry Tamminen, secretary of the California Environmental Protection Agency. "People won't see immediate benefits from this but they need to understand that their children will."
Auto industry officials argued vehemently against the regulations on three points that the board did not have the authority to adopt the regulations, that they could not be met by current technology and that they unfairly targeted California, which produces less than 1 percent of the world's greenhouse gases.
The board said its research had found that the regulations would result in vehicle price increases that would top out at about $1,000 more per vehicle by 2016. The auto industry has estimated the increase at about $3,000, but the board's staff said that number was exaggerated.
The industry will have until 2009 to begin introducing cleaner technology, and will have until 2016 to meet the new exhaust standards. The proposals would require automakers to reduce emissions by using such technological innovations as better air conditioners, more efficient transmissions and smaller engines.
It's worth doing the math here. First, remember that global warming is a global problem, so we have to ask what the percentage reduction in total greenhouse emissions is. As noted above, California only contributes about 1% of the total greenhouse emissions. According to DOE, transportation contributes about a third of the total US greenhouse gas emissions. So, if we reduce emissions from vehicles by 25%, the total improvement is:
.01 * .33 * .25 = .0008 = .08%.
So, the direct benefit of this legislation will be at best a .08% improvement in greenhouse emissions. Actually, the situation is worse than that, since it takes about 10 years for the motor vehicle fleet to turn over, so we won't get that benefit for ten years. And, of course, making new cars more expensive--which even the CARB agrees this will do--makes it less attractive to buy new vehicles, which makes the fleet turn over slower.
This move only makes sense if you think that the rest of the US is going to follow California--as they have sometimes done in the past--and enact similar rules or that the automakers will decide they don't want to make two models of vehicles and make all their models California-compliant. Even then, since the US contributes about 20% of greenhouse gases, this would be a modest 1.6% reduction in world greenhouse emissions. This makes it pretty hard to argue for this change on the grounds that it significantly benefits California. On the other hand, if your goal is to make a statement...
One of the downsides of hiking is that the California forests are infested with poison oak. We spent a lot of our trip pushing through brush, so I wasn't that surprised to discover that familiar red, itchy, rash on my arms and legs. Now, I've had poison oak before and know that topical hydrocortisone cream is basically worthless, so I was expecting the traditional 2-week course of oral prednisone. The good news is that the situation has improved: my doctor gave me a prescription for lidex ointment, which works much better than hydrocortisone. It's probably not quite as good as prednisone, and since it doesn't require spending two weeks on oral steroids, that seems like a worthwhile trade.
In the second part of our series, we'll design a workout plan based on our five principles. We're going to have two basic kinds of workout: aerobic cardiovascular (cardio) and anaerobic strength training.
Cardiovascular training is about the ability sustain low to high moderate intensity levels for long periods of time. Think running distances more than a half mile or so, the Tour de France, that kind of thing. The main element in endurance performance is the ability of your body to deliver oxygen to your muscles and your muscles to use it efficiently to do work. The way you train these systems is by doing low to moderate (and sometimes high) intensity for long periods of time.
There are a lot of different forms of cardio training, with equipment requirements ranging from next to nothing to complicated machines. The standards are:
None of these are inherently better than the others in terms of training your crdiovascular system, though some ahve other advantages and disadvantages. Running, for instance, is fairly hard on your joints. It's mostly a matter of taste and you can mix them up. I'll provide more details about how to perform cardio workouts in a future post.
In this context, strength is about your ability to deliver a lot of force (and energy) over a very short period of time. Think lifting heavy objects.
For strength training, there's only really one option: weights. If you don't have a membership in a gym, you'll need to get one. Weights come in two forms: free weights and weight machines. Free weights are the classic weight training equipment: barbells and dumbbells. Weight machines are basically stacks of weights attached to levers, pulleys, etc. letting you simulate a lot of the workouts you can do with free weights and some you can't.
The free weight versus weight machine debate is never-ending, but most people who lift seriously prefer free weights, as do I. The basic argument for free weights is that the need to stabilize the weight contributes to the development of secondary muscles. However, for at least some of the lifts you'll need a spotter, so they work best if you have a training partner. There are also some exercises where machines really dominate free weights. You'll want to find a gym that gives you options.
Our basic training program is simple: 3 cardio workouts and 2 strength workouts a week. In the basic workout, each cardio session will take 40-60 minutes and each weight session will take about an hour. You want to space them out so that that you get the right amount of rest. This means having the weight workouts 2 and 3 days apart. Here's a sample schedule:
There's nothing special about this arrangement. Certainly, you should feel free to put the days off whenever they're convenient for you. One thing you should pay attention to is where the days off are with respect to your lifting and cardio days. If you find that one set of workouts is more tiring than the others, then try to position the days off so that you recover as fully as possible from them.
Here ends the second part. In the next two, I will provide detailed descriptions of my recommended cardio and strength workouts.
Recently, a friend, sensing that I'm a workout freak, requested some advice on a workout plan for "general fitness". I've gotten this kind of question before, so I figured this might be of general interest. This post is the first in a series describing such a workout plan.
Before we start, let's be clear on the goal: it's not to turn you into Dave Scott or Magnus Ver Magnusson. Being a serious athlete takes a lot of training. The intention here is merely to maintain a reasonable level of fitness with a minimum of suffering.
I'm basically assuming that you don't have much of a workout program right now. Maybe you play the occasional game of hoops or go rock climbing once a week, but to a first order your workouts consist of walking to the fridge to get another pint of Ben and Jerry's. If you've already got something more serious going on, you can probably work some of the stuff I'm talking about in, but I'm not telling you to throw that away and do what I say.
In this post, what I want to do is provide a broad overview of the program, which I'll detail in future posts.
Before you start working out, you need a plan. A plan serves two major purposes:
While there are a lot of training programs that are compatible with getting fit, there are also a lot of programs that aren't. If you just wake up every morning and do whatever workout pops into your head, you're a lot more likely to end up with a program (if you can call it that) that is one of the latter than one of the former. Sitting down and designing a training plan based on sound principles lets you design a good program.
The second purpose of having a training program is that it keeps you on track. From time to time, you may find that you don't feel like training that day. This is especially common when you're just getting started or you've got other stuff going on in your life. If you're just going from day to day, it's easy to stop working out--and then hard to start again. If you have a plan that tells you what you need to every day, it's a lot easier to stick with it.
A plan can be complicated, but it doesn't have to be. I plan my workouts a month at a time and the calendar contains every workout I have that month, down to the minute and intensity level. My friend Kevin has a more informal system: he knows more or less what he's got going on every day but calibrates a lot based on how he feels. You certainly don't need to have a plan as detailed as mine, but you should at least know what workouts you have each day and what the purpose of each workout is.
One good way to maintain your plan is to keep records. I like to write down my plan for the month and then keep records of which workouts I did and how they went. This lets you track your progress and readjust your future plans based on that. If you keep long term records it also gives you an opportunity to look back and see what worked for you in the past. This can be particularly useful if you get sick or injured and really fall off the program, because then you know how to build yourself back up again.
In our program, we'll focus on developing a program for one week of training at a time.
There are a lot of different kinds of physical fitness: strength, muscular endurance, flexibility, cardiovascular fitness, etc. Most sports depend on all of these to one extent or another--though a few sports, such as long distance running, are pretty specific.
However, the best way to improve performance in any particular dimension is to do a workout that focuses on that and that alone. If you are trying to improve multiple dimensions, you should train them independently. If you're doing a sport that depends on multiple kinds of fitness, you want to mostly train them independently and then do sport-specific workouts than let you practice your skills.
These principles apply in the small scale as well as the large. Say you're a road runner (10ks, marathons, etc.). You do what's called multipace training: some days you do easy distance; some days you do hard intervals; some days you do speed. In the same way, boxers do roadwork to train their cardiovascular systems, lift to build strength, and drills and sparring to develop their skills.
The more serious you are (and the more workouts you're willing to do per week) the more different kinds of workout you can fit in and the more specific you can make your training. In our program, we'll be doing two basic kinds of trianing: aerobic cardiovascular and strength, with some finer divisions intended for people interested in a more advanced program.
The basic principle of training is overload and adaptation. Whatever your current fitness level, the stimulus that forces your body to improve is overload: giving it a training load that is greater than it is accustomed to. Your body adapts to this stress and as your fitness improves, you become able to tolerate training loads that you previously could not.
They key to making the overload/adaptation principle work is slowly building up your training load and giving yourself time to adapt. The time scale is fairly long. Olympic marathoner Frank Shorter used to talk about the "two week/two month" rule. It takes about two weeks for you to get comfortable with an increased training load and two months to fully adapt.
In our program, we'll be careful to build up slowly, making sure that you're getting some training stimulus, but also giving you time to adapt.
The flip side of the overload/adaptation principle is the need for rest. While you need overload to stimulate adaptation, you need rest to give the adaptation time to happen. If you just keep training and never give yourself time to rest, you won't get any better. You'll just get end up overtrained and probably sick.
In order to get proper rest, you want to space your workouts out over the course of the week. In general, it's best to put rest days after your hard workouts. People who use heavier training loads, training every day or multiple times a day, will generally try to use the hard/easy principle, placing easy days in between hard days. Even then, they generally try to take one day of complete rest every week or every two weeks.
The training program we'll be developing has 5 workouts a week with two rest days, spaced out to provide maximum recovery.
The best workout plan in the world doesn't do any good if you don't execute on it. So, the final principle is that you need to follow through. If you've got a workout scheduled for the day, you should try to complete it. If for some reason you can't make a workout one day you should try to make it up if you can.
Obviously, there's some tension between this principle and the rest principle. There are basically three reasons why you should skip workouts:
Obviously, it's a judgment call whether you should skip a workout. It's especially difficult to determine whether you're just a little tired and feeling lazy or whether you're actually overtrained. This kind of judgment comes with time, but I'll try to provide some guidance in future posts.
Here ends the first part. In the next, I will provide an overview of the training program and a sample schedule.
Acknowledgement: thanks to Kevin Dick for helping me frame some of the ideas (in particular the principles) enumerated here.
Colby Cosh has an interesting article about European hockey, where apparently advertising on uniforms is standard, as shown in the picture below:
Apparently soccer also has this.
I've previously posted about how advertising on cycling jerseys affects tactics. I know next to nothing about hockey or soccer (hockey without the ice, right?) but I wonder if advertising changes tactics. As we've seen, in cycling it encourages various kinds of showboating by the individual riders because it focuses attention on their sponsors and away from the sponsors of the other teams. In a hockey game, by contrast, I'd imagine that players from both teams are on screen most of the time so showboating is less useful.
What would change behavior a lot, I expect, is to allow individuals to contract for their own sponsorships. that would give them a lot of incentive to get screen time for themselves rather than help their teams. But that's not how it works, right?
Sacramento -- Aiding the industry that helped him gain worldwide fame, Gov. Arnold Schwarzenegger signed legislation Tuesday aimed at discouraging online piracy by requiring anyone disseminating movies or music on the Internet to disclose their e-mail address.
California file sharers who trade songs or films without providing an e- mail address will be guilty of a misdemeanor, under the first-in-the-nation measure that could make it easier for law enforcement to track down people who illegally download copyrighted material.
The bill is the latest attempt by film and music trade associations to combat the hard-to-police use of file-sharing software.
Annoying, I suppose, but I doubt it will have any real impact on anything. It's not like it's hard to get a Hotmail/Yahoo/Lycos account and they're not really much easier to tie to your real world identity than whatever information is available from the filesharing traffic.
Courtesy of Stupid Security comes the story of a flight that was cancelled because a passenger found a note in Farsi in an in-flight magazine. I hope the folks at the TSA don't hear about this or they might start confiscating pens and paper at the security checkpoints.
Lisa and I spent the weekend in Los Padres National Forest again.
|Driving distance||150 miles (each way)|
|Walking distance||23 miles (loop)|
|Max altitude||4757 feet (from TopoUSA)|
|Total climbing||7089 feet (Lisa says "Aargh!")|
|Scenery rating||9 (high meadow, river valley)|
|Weather||Cool and mostly shaded, rainy on day 3|
|Water availability||Yes. You're along a river most of the time|
|Special hazards||Some trails hard to follow|
|Cost||Zero! (the USFS has waived the adventure pass)|
|Permits required||Campfire permit|
Lisa and I left Palo Alto about 10 AM and arrived at the China Camp campground and trailhead about 2:30 and were on the trail by 3. Warning: the last 5 or so miles of Tassajara Road on the way to China Camp are a seriously rutted and potholed dirt road. We were able to make it ok in Lisa's aspire, but a high clearance vehicle with a better suspension would be nice here.
The first mile and half of the route is a road descending from 4300 feet to about 3600 feet. I haven't quite figured out how to make TopoUSA route over both road and trail, so just trust me that it's the gray line from the red dot (labaled Tassajara Rd) to the black dot labelled Start.
We descended the rest of the first day and made camp at the Clover Basin camp (the first marker on the profile), which is a small campsite looking over a stream. There are two difficult bits here. The first is that Nason's Cabin is supposed to be located at a saddle point about 4 miles in. All we saw was a fence in a field with two paths going out of it. We followed the wrong trail for a while before Lisa worked out what the right one was. Towards the end of the day you join up with the river and there are some tricky bits following it. Clover Basin camp is generally pretty nice, but someone had left a whole bunch of old clothes and stuff at the site. Pack out your crap, people!
The second day follows a series of creeks and streams almost the entire way. There are about 25 stream crossings, so if you plan on hiking this route bring waterproof boots. I had waxed my boots just prior to leaving so my feet stayed pretty dry. It was sometime around the fifth stream crossing that the sole of my boot (recently resoled!) decided to separate from the boot proper. I ended up having to make an emergency repair with a nylon strap, which worked fine after a little trial and error:
The second half of the day we hiked up to Pine Valley, which is located on a high meadow. The trail starts to get pretty overgrown here and we had to beat our way through a bunch of brush, which made for slow going. We camped in what looked like a slightly ad hoc campground (the second marker in the profile) right in a stand of pines in the center of the meadow. The view was beautiful (or so Lisa tells me) but by 7:00 it had started to get really cold and windy so we ended up eating dinner all bundled up and turning in at 8:00 or so.
We were up early the next morning for the hike out and were about halfway through breaking camp when we felt the first drops of rain. We rushed through packing the tent and had only made it about a half mile before it was seriously coming down. We were actually pretty prepared for the rain since we had pack covers, waterproof jackets, and warm clothes, but since the next 2.5 miles or so consisted of pushing our way through wet brush, so by the time we got to the first summit we were pretty much completely soaked anyway. After another two miles of that we were definitely ready to put on some dry clothes and sit in a heated car.
If you're doing performance engineering, the first thing you need to know is Amdahl's Law's_law. The amount of speedup you can get by optimizing a specific operation in a system is limited by the fraction of overall system cost consumed by that operation. Specifically, the speedup is:
|speedup =||Execution timeold
(1 - Fractionenhanced) + (Fractionenhanced / Speedupenhanced)
The key point here is that if operation X consumes Y% of your program's run time, no matter how fast you make X, you can't speed the program up by more than Y%.
One specific instance where I happened to run into this lately is SSL performance. It's well known that the dominant cost of doing SSL on a web server is the RSA private key operation. Obviously, anything that improves the performance of this operation will also improve the performance of your SSL server, which is why cryptographic hardware that accelerates RSA is popular, and why there's interest in algorithms that are faster than RSA.
Unfortunately, it's easy to overestimate the improvement from this kind of optimization. It's pretty common to see assertions like this one that because algorithm X is Y times faster than RSA, SSL with algorithm X will also be Y times faster, which simply isn't true. In fact, the performance limitations are surprisingly tight: in 2002, Coarfa et al. published measurements indicating that even if the RSA operation were completely eliminated, in realistic scenarios1 the performance improvement is limited to around a factor of two. Amdahl's Law strikes again!
1 Well, at least semi-realistic. The workloads were synthetic, but the assumptions they made were plausible. The key factor controlling how much improvement you get is how much session reuse/resumption goes on. The more session resumption there is, the smaller the benefit of accelerators. However, even with no resumption, the improvement is less than a factor of 4.
Lisa and I went camping this weekend in Los Padres National Forest (more on this later). Some of the trails are pretty badly marked and we ended up taking a few wrong turns and then having to backtrack. True, we had both a topo map and a GPS, which helps some, but the distance between the right and wrong trails can be as short as a few hundred feet and even when we took the right path, there were a few times when we had to go several miles to the next major landmark before we were absolutely sure we had made the right choice.
No doubt with a lot of practice and calculation, we would get a lot better--after all, people do orienteering without any GPS at all--but until then I prefer a technical fix: a GPS with mapping capability. I use TopoUSA for my home mapping software, so what I'm looking for is a GPS receiver that I can download maps (including the trails we plan to follow) to, thus allowing me to navigate with only the GPS most of the time.
Do any EG readers have map-capable GPSes they would like to recommend?
Wired (and other media outlets) are reporting that it's trivial to open a Kryptonite bike lock using a Bic pen barrel. See "here": for a video. Ed Felten has an interesting post about the implications of this result for the debate on disclosure.
Here's the part that I find weird about all this. Based on the BikeForums postings, it appears that all you need to do is get the pen barrel into the key channel (this may require a very small amount of cutting the barrel) and then the thing just turns and unlocks. Is it really possible that anything that will grab the lock barrel will open the lock? Wasn't this like a pretty obvious flaw at design time?
Slate has an Explainer of what an assault weapon is:
For starters, consumers can now purchase one of the 19 firearms specifically outlawed by the 1994 ban, including such action-movie staples as the Uzi, the TEC-9, the Kalashnikov, and the Street Sweeper shotgun. The law's authors had to be as precise as possible in crafting the ban, since the phrase "assault weapon" isn't really part of the gun-making vocabulary. Rather, it's a catchall term that gun-control advocates define as covering any firearm designed for rapidly firing at human targets from close range. The 19 guns called out in the ban are all semiautomatic in nature: They can eject spent shell casings and chamber the next bullet without human intervention, but only one round is fired per squeeze of the trigger. Not coincidentally, the verboten 19 were largely the sorts of high-profile guns that even gun neophytes have heard of, either through watching the nightly news or by reading Tom Clancy novels.
On top of the Big 19, the ban also included a few formulas for forbidding less well-known armaments. A semiautomatic rifle was considered an illicit assault weapon, for example, if it featured a detachable magazine, as well as at least two of the following five attributes: a folding or telescopic stock; a conspicuous pistol grip; a bayonet mount; a flash suppressor or threaded barrel (i.e., a barrel that can accommodate a flash suppressor); or a grenade launcher. The checklist for semiautomatic pistols includes guns weighing more than 50 ounces when unloaded, and those featuring a "shroud" on the barrel to prevent a shooter's non-trigger hand from being burned.
This isn't really that great a summary, though.
The assault weapons ban has always been a sham. Too bad this article didn't make that clear.
According to the Register, the UK Consumer's Association has asked the Office of Fair Trading to investigate iTunes's pricing practices. The table of prices is below:
|Country||Price||Price in Dollars (as of today)|
Note that everyone in Europe and the UK is paying more than we do here in the US.
Obviously, there's nothing unusual about having domestic pricing and it's clear that they're trying to choose prices that are near psychological anchor points. From the prices alone, it's not obvious that iTunes is engaging in price discrimination. After all, exchange rates fluctuate and it wasn't that long ago that €.99 was less than $.99.
However, there is one piece of evidence that suggests that the price discrimination is intentional: iTunes doesn't let people in the UK buy from the European service. You need to have an address in Europe to do so.
Digital Journalist has an exhibit of 100 Photographs that Changed the World (+ Marginal Revolution). I would also include Rosalind Franklin's DNA X-ray diffraction pattern which directly lead to Crick and Watson's structure.
Any other candidates?
UPDATE: Fixed broken link above. Original link was the victim of a cut-and-paste mistake.
Tomorrow's Science Times has a Q&A about male pattern baldness. Turns out that it's not X-linked, as urban legend has it:
It was long thought that inheritance of male-pattern baldness was a fairly simple matter involving a single unidentified dominant gene, and many assumed it was passed on by the mother on the X chromosome. A copy of the form of the gene that predisposes to baldness was thought to mean a man would lose hair in the typical pattern. The gene was believed to have reduced effects or to be recessive in women, who needed two copies for baldness.
However, this view was based on a relatively small study of 22 families done in 1916, and more recent research suggests a more complex situattion, one involving multiple genes, not necessarily passed on in the maternal line, and one with an additive effect with each gene involved.
A larger Australian study reported in The Journal of Investigative Dermatology did not identify a candidate gene as the culprit and in fact found no clear pattern of inheritance but noted that there was a relatively strong correspondence between baldness tendencies in fathers and sons.
On the other hand, in my family I think it's pretty clear where the fault lies. When my father was my age, he had a pretty much full head of hair. By contrast, one of my maternal uncles went bald in his early-mid twenties, which is when I started losing my hair, before I mounted my pre-emptive strike.
For the past two years or so, John Gilmore has been fighting a lonely battle against the requirement that you show ID to travel by air. At the moment, the airlines claim they're required to ask for it but that the law is secret:
That's what John Gilmore wanted to know. At San Francisco's airport, just like the rest of the country's airports, there was a sign that began "A Notice From the Federal Aviation Administration" and includes the sentence "passengers must present identification upon initial check-in.
John worked his way up the bureaucratic chain and was eventually told by United Airlines that there were security directives that mandated the showing of ID, but that he couldn't see them. These secret directives, issued by the Transportation Security Administration, are revised as often as weekly, and are transmitted orally rather than in writing. To make things even more confusing, these orally transmitted secret rules change depending on the airport.
Gilmore's original case was dismissed by the U§ District Court (it said that it didn't have jurisdiction) and so he appealed to the Ninth Circuit, which is where the case is now.
The government's position seems to be that the text of any law requiring that ID be shown--or even the existence of such a law--is a matter of national security and so they asked to submit their brief under seal:
Moreover, defendants are precluded by federal statute and regulation from disclosing the existence and content of any such security directive. Under 49 U.S.C. § 114 (s)(1)(c), "the Under Secretary [of Transportation for Security] shall prescribe regulations prohibiting the disclosure of information obtained or developed in carrying out security . . . if the Under Secretary decides that disclosing the information would be detrimentatl to the security of transportion." 49 U.S.C. § 114 (s)(1)(c). Pursuant to that authority, the Under Secretary has defined a set of information known as "SSI" or "sensitive security information" (see 49 C.F.R. § 1520.3), and has directed that such information shall not be disclosed except in certain limited circumstances not applicable here. 49 C.F.R. § 1520.9(a)(a) ("A covered person must . . . disclose . . . SSI only to covered persons who have a need to know, unless otherwise authorized in writing by TSA." The Under Secretary has defined SSI to include "[a]ny aircraft operator or airport operator security program" and "[a]ny Security directive or order . . . [i]ssued by TSA." 49 C.F.R. § 1520.5(b)(1)(i), (b)(2)(i). By regulation, aircraft operators must also "®estrict the distribution, disclosure, and availability of information contained in the security program to persons with a need-to-know." 49 C.F.R. § 1544.103(b)(4). Accordingly, even if plaintiff's allegations about the existence and content of the alleged security directive were correct, defendants were prohibited by statute and regulation from openly disclosing those facts before the district court, plaintiff, or plaintiff's counsel.
Today, the circuit court rejected this motion, so it appears that the government is going to refuse to provide any supporting material and stipulate that the order exists.
Forget about whether you think that the ID requirement is a good idea or not and just look at the situation tactically. What purpose is being served by the government's refusal to provide a regulation? The government lawyers say they can't do so, but if you read carefully it seems pretty clear that the Under Secretary could instruct them to, so that doesn't seem like a real obstacle.
It seems to me that there are at least four major possibilities:
I don't think we know enough yet to distinguish these...
UPDATE: fixed "district" court above to read "circuit".
BBC reports that the Serbian government has reversed a decision (not that I'd ever heard of it) to ban the teaching of evolution in schools:
"I have come here to confirm Charles Darwin is still alive," said deputy education minister Milan Brdar.
His boss, Ljiljana Colic, who had announced the controversial policy, had gone "away on business", he said.
She had proposed banning the evolution theory this school year, until creationism could be taught alongside.
Both Darwin's theory of natural selection and the Old Testament view on the beginning of life were equally dogmatic, the minister had said.
After a deluge of protest from scientists, teachers and opposition parties, Serbian Prime Minister Vojislav Kostunica called Ms Colic in for a meeting.
They agreed to drop the move, Mr Brdar said.
I hadn't realized that creationism had substantial pull outside the United States, since in the US it's tightly tied to evangelical Christianity. Apparently in Serbia it's the Eastern Orthodox Church which is behind it. I wonder what gets taught in schools in the religious Muslim countries?
I'll be speaking at the DIMACS Workshop on Cryptography: Theory Meets Practice being held at Rutgers on Oct 14-15.
Title: What's the worst that could happen?
Communications security systems depend completely on the security of the underlying cryptographic primitives on which they are built. However, these primitives are always imperfect, sometimes badly flawed and occasionally completely broken. In this talk, we will examine the impact on important communication security protocols if attacks--even partial ones--were to be found on our core cryptographic primitives.
The rest of the program looks pretty interesting too.
You may have heard about the South Korean admission that they enriched some uranium back in 2000. One interesting feature of the reports is that the South Koreans used laser enrichment rather than one of the more standard methods. This week's Science has an interesting article describing how the technique, called AVLIS, works:
The method in question is known as atomic vapor-laser isotope separation (AVLIS). AVLIS exploits a subtle difference in how uranium-235 and uranium-238 absorb light. Because the two atoms have different masses, they absorb very slightly different colors of light. By shining a laser of precisely the right color on a beam of mixed-isotope uranium vapor, scientists can induce the uranium-235 in the beam to absorb a photon of light and fly in one direction while the uranium-238 in the beam remains unaffected. That's the theory, anyway.
In practice, though, AVLIS hasn't proven useful for separating uranium on a large scale. "There are no commercial programs" that use lasers to separate uranium isotopes, says Thomas Cochran of the Natural Resources Defense Council in Washington, D.C. A diplomat who is knowledgeable about nuclear proliferation issues says the countries that have tried it concluded "it was too expensive; you could not produce enough [enriched uranium] quickly." In 1999, the United States killed its own AVLIS program, developed at Lawrence Livermore National Laboratory in California and run by a private firm based in Maryland.
Pretty clever, really.
Science is reporting that a leading HIV vaccine candidate basically doesn't work:
The 4-year-old study, funded by the New York City-based International AIDS Vaccine Initiative (IAVI), is taking place in five countries, but these preliminary results are from the United Kingdom, Kenya, and Uganda. Although many AIDS vaccines have focused on triggering production of antibodies that prevent HIV from infecting cells, this trial tested whether two vaccines in combination could stimulate the so-called cellular arm of the immune system, which clears cells that the virus manages to infect. The study built on provocative evidence from HIV-exposed but uninfected sex workers in Nairobi and the Gambia. McMichael and other researchers found that these subjects had developed cellular immune responses to the virus (Science, 23 June 2000, p. 2165).
The closely followed study has broad implications because several other research groups are pursuing similar approaches. Both vaccines rely on harmless vectors to shuttle an HIV gene (gag) and other small pieces of the virus into the body. The "priming" vaccine splices the viral components into a ring of bacterial DNA, and the researchers follow it with a "boost" that delivers the same HIV ingredients by means of an experimental smallpox vaccine called modified vaccinia Ankara (MVA).
The McMichael team measured the ability of the prime-boost vaccination to turn up production of the biochemical messenger interferon in response to HIV, an indicator that the immune system has launched a cellular attack against the virus. The negative, preliminary results led IAVI to scotch plans to expand the MVA/DNA trials to other countries, but the researchers will complete those that are under way.
McMichael says their results may be disappointing in part because the team was very stringent in how it defined a positive interferon response. But he also suspects that the DNA prime, which works well in mouse experiments, didn't do its job. "I think DNA is a poor primer in humans," says McMichael, who notes that it has performed badly in other human studies. Yet there's no denying the new data call into question the worth of MVA. "Is this the death knell for all MVAs?" asks Cornell University's John Moore, a member of NIAID's AIDS Vaccine Research Working Group. "If other MVAs are no more immunogenic than McMichael's, this has major strategic impact.
Not very encouraging.
The government's attempt to block Oracle's bid for PeopleSoft has failed. For reference, Oracle has bid $21/share and PeopleSoft is currently at $17.95 and hasn't been above $20 since March of '04, making $21 look like a pretty good deal at this stage in the game.
The other night I had dinner with my friend Cullen who told me an interesting story about his days at a startup. When they went to hire a receptionist, a number of the candidates were attractive young women who told him that while the job seemed good and the pay was fine, their primary motivation was to meet young, single engineers who would be good marriage material. Cullen told them to walk down the hall and make their own assessment.
If this is a common phenomenon, it suggests an interesting form of signaling and a non-sexist reason to hire an attractive receptionist: it's an indication that that you're a dynamic company with young, smart, engineers--unless, of course, it signals that your engineers look particularly clueless and desperate.
Credit: The signaling observation came out of a discussion with Kevin Dick
One issue that didn't get brought out by Matthew and Gregg's posts was that the Krueger and Dale did find a good predictor of future income: college tuition.
The point here is that you shouldn't worry so much about whether or not you (or your child) gets in to Harvard (or some other elite school, I'll just use "Harvard" as shorthand from now, like Easterbrook, on with no slight intended to those who were admitted to other highly selective institutions). The proper conclusion to draw from the Krueger and Dale data, however, is rather different. Their research indicates that there may be no good reason to attend Harvard if you can, as will usually be the case, get a more attractive financial aid package from a less selective school or else simply find a lower tuition at a public university. Their research most emphatically does not support the conclusion that whether or not you can get admitted to a highly selective college matters. On the contrary, the research indicates that the methods used by the admissions officers at these schools are rather good at identifying persons who are likely to achieve high incomes later in life. Going to Harvard, then, may have no particular value, but it's still of a great deal of interest to you whether or not you.
Actually, it's a little more complicated than this. Let's assume that your actual quality value is Q. Now, neither you nor the college admissions staff can actually measure Q. Instead, you both have to estimate it. Now, the admissions company makes their estimate based on some set of observable performance variables Oi (SATs, extracurricular activities, grades, etc.) You make your estimate based on those variables plus some unobservable variables Ui (how hard you worked in any given class, your estimate of how smart you are compared to your friends, etc.) In addition, you use different algorithms to compute your estimates. So, this leaves us with:
|Qy = Fy(Oi,Ui)||your estimate of Q|
|Qa = Fa(Oi)||the college's estimate of Q|
Now, the important thing to realize is that Oi are generally not used by employers to assess the quality of applicants. They use other measures, like college grades, recommendations, etc., which at best are coupled to the other Oi values through Q. But Q is the real predictor of your ultimate earning power!
So, say you know Qa, what good does it do you? Well, it lets you update your estimate Qy of your Q value, on the theory that the admissions committee's algorithms may be more accurate--or at least have uncorrelated errors. To sharpen this point, imagine there was an easy way to improve one of your observables, for instance by bribing a teacher to get a higher grade. This would increase the college's estimate of Qa, but it shouldn't lead you to update your Q estimate, because you know that that Qa doesn't accurately reflect Q.
So, it's true that getting into a good school is a good signal that you have qualities that are likely to stand you in good stead later in life. However, that doesn't mean that doing things that are more likely to get you into that college improves your future prospects.
The researchers Alan Krueger and Stacy Berg Dale began investigating this question, and in 1999 produced a study that dropped a bomb on the notion of elite-college attendance as essential to success later in life. Krueger, a Princeton economist, and Dale, affiliated with the Andrew Mellon Foundation, began by comparing students who entered Ivy League and similar schools in 1976 with students who entered less prestigious colleges the same year. They found, for instance, that by 1995 Yale graduates were earning 30 percent more than Tulane graduates, which seemed to support the assumption that attending an elite college smoothes one's path in life.
But maybe the kids who got into Yale were simply more talented or hardworking than those who got into Tulane. To adjust for this, Krueger and Dale studied what happened to students who were accepted at an Ivy or a similar institution, but chose instead to attend a less sexy, "moderately selective" school. It turned out that such students had, on average, the same income twenty years later as graduates of the elite colleges. Krueger and Dale found that for students bright enough to win admission to a top school, later income "varied little, no matter which type of college they attended." In other words, the student, not the school, was responsible for the success.
This is quite an interesting result, because it doesn't match up with either of the standard hypothesis for the education premium--the fact that people with college educations make more money.
The first hypothesis, which I'll call training is that education actually teaches you stuff that improves your job performance. So, for instance, if you're going to be a computer programmer it might be helpful to take some classes in computer programming. Naturally, then, people who get college degrees will tend to have more marketable skills and be able to make money.
The second hypothesis, called signaling, is that education is a way of demonstrating your abilities. In many jobs, what you want is for the employee to be smart and dedicated. You need to be smart and dedicated to get a college degree and so getting one is a good way of showing that to your employer. Thus, an employer will be willing to pay more for people who have college degrees even if their education didn't give them any relevant skills.
You can apply similar reasoning to the salary premium commanded by graduates of elite universities: in a training view, going to Harvard actually teaches you stuff which will make you more effective later in life--stuff that you wouldn't have learned going to, say University of Connecticut. In the screening view, the people who get into Harvard are smarter than the people who get into UConn--and employers know this--so they're more interested in hiring them. In essence they can use admission to an elite college as a proxy for underlying applicant quality, without having to measure it directly.
This result doesn't neatly fit into either view. Obviously, if people who attend an elite school don't make more money than those who are accepted but don't attend, that really damages the training hypothesis. If elite schools teach you stuff, then the people who actually attend them should make more money. However, this data it doesn't really support the signaling hypothesis either.
Say that each employee has some underlying quality value Q and that Q is correlated to later earnings. Now, the fact that people who are accepted to elite schools tend to earn more is easily explained by arguing that the admissions process attempts to measure Q and that people with high Q tend to get into the elite schools. No problem there: it looks like the admissions process is pretty efficient. But if the process is so efficient, then why don't employers care about it? Everyone who attended an elite school was necessarily accepted and therefore is likely to have a high Q value. Now, to be sure, some of the people who didn't go to elite schools were accepted, but most weren't. So, people who attended elite schools are likely to have high Q values. Why aren't employers willing to pay a premium for high-Q employees?
One possibility is that they are, but that the effect fades awy. Reading the original paper, it appears that the salaries used are from people who have been out of school for some time. So, it may be the case that starting salaries are higher but that long-term salaries are more affected by actual performance and therefore directly measure Q, rather than relying on signalling from college attendence.
This month's Network Magazine cover story is about the evolution of the Internet Protocols. At the end of the article, there's a Q&A with Fred Baker (on TCP/IP), Eric Allman (on Spam) and myself (on Security). There's nothing here that will be too surprising to EG readers, though.
They have announced three competitions:
The claimed goal is to demonstrate (or disprove) feasibility of a space elevator by 2010. I can see how we would demonstrate feasibility, but I'm not sure I understand how it would disprove it. Presumably a lot of the uncertainty is in the materials science for the tether, and that's subject to scientific discoveries that don't necessarily occur on any particular timetable.
A while back Eugene Volokh mentioned the ExpressO law review submission service. The idea is that you send your article to ExpressO and then it will submit to the law reviews you select. When I tell colleagues about this tool, they're invariably surprised, not that it exists, but that there would even be a demand. You see, in CS (and I understand many other fields) submitting to multiple venues simultaneously is verboten. Indeed, it's grounds for immediate rejection without consideration of the merits of your paper. And yet in law, not only is it considered normal, there is a tool to facilitate it!
The interesting question, then, is why the difference? I suspect it's something simple: there are a very small number of CS publishing venues. For instance, in security there are probably less than 10 "good" venues. By contrast, there are upwards of 400 law reviews, so if everyone just submitted to the best ones, the 399th ranked one wouldn't get any submissions at all. Of course, this causes an enormous amount of wasted review energy, but unlike CS, law reviews are staffed by students, which is no doubt why there are so many.
Of course, this leaves the question of why legal publishing is run by students...
Here's an interesting article about hacker attacks on WalMart gift cards:
he same thing happened to Carol Kent and her husband with a $25 card at the Wal-Mart in Puyallup.
Carol: "She said 'I'm sorry, but there's a zero balance on this.' And we're like, 'What?!' She ran it again and she said, 'No, I'm sorry. It's already been cashed out.' "
Carol's shopping card was purchased in Olympia, and days later, cashed out by a stranger at the Wal-Mart in Chehalis even though Carol still had the card.
"Here's my receipt," Carol points to the shopping card notation at the bottom which reads: "Shop card reception 0.00"
In Tami's case, her receipt shows the $150.00 card was activated at 11:32 in the morning, then cashed out three hours later in a another state!
"At a store in California," Tami explained. "He (the Wal-Mart employee) wasn't sure how it was being done, but he told me it had happened several times through that same store in California."
It seems to me that this must be some kind of inside job. As I understand it, these cards just have unique numbers which are used as database lookup keys back at WalMart central. When you buy a card, they activate it in the central database. Purchases just deprecate the database balance. There are obviously lots of ways to attack this kind of system. For instance, a cashier could simply falsely activate a card and then start spending with it. Or, obviously, someone with access to the central database would be able to mount this kind of attack.
I finally got around to reading the August 13th Science issue on the Hydrogen Economy. Generally, the articles are pretty skeptical.
One of the more interesting artices was a comparison between hybrid and fuel cell cars by Nurettin Demirdöven and John Deutch. One of the most likely ways to use hydrogen to power a vehicle is to use it in a fuel cell. The authors compare the end-to-end efficiency of hydrogen fuel cell-powered cars to ordinary gasoline hybrids like the Toyota Prius and get some interesting results.
The situation is roughly as follows: fuel cells are a lot more efficient than internal combustion engines, but you most likely will not be able to consume gasoline directly. Instead, you need to convert the gasoline to hydrogen using a process called reformation. The figure below (taken from the paper), shows the estimated efficiency losses at each stage of the energy delivery train. The three vehicles being compared are:
Both the HICE and AFC vehicles are much more efficient than the conventional ICE vehicle, but the HICE and AFC vehicles are very close in terms of efficiency because the engine loss and the loss in the reformer mostly wash out.
A lot of hydrogen economy advocates think of this kind of reformer + fuel cell vehicle as a transition strategy. In the envisioned hydrogen future, the hydrogen production will happen more centrally and gas stations will be replaced with hydrogen filling stations. But even the best current reformation technology is only about 85% efficient (as opposed to the 80% assumed in the model shown above) so even in this scenario the advantage of fuel cells over hybrids wouldn't be that great.
This afternoon, I had the joy of being selected for airport "secondary screening", meaning I got to sit in a chair for 10-15 minutes while some TSA employee ostentatiously paws through all your stuff. As should be obvious at this point, this is pure security theater. I had any number of items which could be carrying explosives or weapons, but you would have needed to look inside to find out. If they didn't show up on the X-ray, I highly doubt that some uneducated fondling will uncover the truth.
While I was standing in line, the TSA employees told me that I was probably selected because I'd bought a one-way ticket. I'm amazed to hear that they're actually using such a stupid selection criterion. Surely, even a small amount of research would turn this up, allowing a terrorist to buy a round-trip ticket. Look, if you're going to select people for a more thorough search, you need to use criteria that are either secret or difficult to change. Otherwise, any sensible terrorist will just change them.
I'm away on business tonight and accordingly am getting to experience the joy of remote Internet access. I currently have two options:
Usually, these work pretty well, but this trip I've managed to run into two annoying problems.
The first problem was with the hotel network last night. I SSHed into my home machine. During a few big file transfers, I've gotten SSH MAC errors followed by connection failure. I suppose I could have been under active attack, but realistically I suspect that the hotel network is doing some kind of NAT-oriented connection surgery that's breaking the SSH packets. Seems to be OK tonight, though. Strange.
The second problem is clear, but even more baffling. I use Eudora on my Treo to read my e-mail via POP over SSL. This worked well, until June 8 when the self-signed cert that my ISP used expired. Now, Eudora prompts me to see if I want to continue. If I click "yes", the connection proceeds and I get my mail. The problem happens when I go to send mail. For some inexplicable reason, the SMTP module doesn't get the message and insists on prompting me again. To make matters worse, it can't seem to remember my affirmative response and just keeps prompting me every time I try to send, so while I can read your mail, I'm completely unable to reply! So, if you send me mail tomorrow and you don't get a response, now you know why.
The Daily Recycler has dug up a video of Hastert implying that George Soros gets his money from drug cartels. Worth watching. You might have thought that this was one of those situations where someone just made a slip of the tongue and didn't want to back down, but no, it looks fairly calculated.
E-Voting scourge Bev Harris has discovered a pretty serious and scary security problem in Diebold's central vote counting program:
Issue: Manipulation technique found in the Diebold central tabulator -- 1,000 of these systems are in place, and they count up to two million votes at a time.
By entering a 2-digit code in a hidden location, a second set of votes is created. This set of votes can be changed, so that it no longer matches the correct votes. The voting system will then read the totals from the bogus vote set. It takes only seconds to change the votes, and to date not a single location in the U.S. has implemented security measures to fully mitigate the risks.
This program is not "stupidity" or sloppiness. It was designed and tested over a series of a dozen version adjustments.
You can't really call this a vulnerability. It's more like a back door.
Apparently this was originally announced July 3, 2003 (though embarassingly, I seem to have missed the announcement). In the meantime, according to Harris, it's been basically ignored.
I'm not so concerned about the hole--though it is very serious--itself, but about the social problems it implies:
I expect voting software to have problems--it's software, and software has bugs. But if we're going to claim that it's safe then the people writing and operating it need to be alert to those problems and responsive when they're disclosed. That doesn't seem to be the case here.