November 30, 2003

The risk of an influenza pandemic

In an ordinary year, influenza might kill 36,000 or so people in the US yearly [*]. A real influenza pandemic would undoubtably be much worse. The 1918-1919 pandemic killed an estimated 20-40 million people. [*], so it's something to be avoided if at all possible.

We haven't had a major pandemic since 1968, so it's tempting to think that we've got things under control. Unfortunately, the experience of SARS combined with Webby and Webster's article in the November 28th issue of Science, leaves me with the distinct feeling that our current influenza strategy is based as much on hope as prevention.

Webby and Webster cite a number of problems:

  1. There are large animal reservoirs of new flu strains. Each year we get new set of strains migrating into the human population. That's why there's a new vaccine every year.
  2. It takes roughly six months to create a new influenza vaccine. We have strategies for faster production but they're not in current use.
  3. A number of the strains of greatest concern won't grow in the eggs used to manufacture the vaccine. Again, we have research techniques for making vaccines with these strains, but they're not currently in use.
  4. We don't have enough antiviral drugs to deal with a pandemic.

None of this is particularly good news.

Posted by ekr at 02:46 PM | Comments (114) | TrackBack

November 29, 2003

Partially computerized voting

In my previous post, I asked whether we could design a system that was both objective and accurate. To tell you the truth, I don't know. However, here's one possibility, which I've seen suggested by someone else, but I can't remember where. As I mentioned previously, optical scan systems are very accurate, but they still have the problem that one could potentially go back and inspect people's hand-marked ballots. So, what happens if we combine DRE with optical scan systems?

The way that this would work is that you'd have a computerized voting machine. However, instead of actually recording the votes, it simply prints out a form which can be fed into the optical scan system. The voter then has an opportunity to verify that the card is correct and can either discard it or feed it into the optical scanner.

This system has a number of key advantages, especially over conventional DRE systems:

  1. It's objective. Since the cards are marked by computer, you can't go back and try to figure out what voter intention was. As I argued in my previous post, this is arguably a good thing.
  2. It's voter-verifiable. The voter can see the voting card and verify for himself that the computerized system didn't cheat.
  3. It's auditable. Since all the cards are captured, you can re-audit the election just as you would with an ordinary optical scan system.
  4. It doesn't expand the trust perimeter much. The only trusted component in the system is the scanner--because people can verify that the computer didn't cheat. However, we already have a scanner infrastructure in place, so we don't have to build a new trusted system. And, of course, the scanners are much simpler than DRE machines and so easier to verify.
  5. People don't have to use the computer. Such a system would still allow hand-marked ballots, thus making transitions easy (though of course the hand-marked ballots wouldn't be as objectively verifiable.)

As I said, this idea isn't original to me. I just can't remember who originally proposed it.

Posted by ekr at 02:55 PM | Comments (57) | TrackBack

What's good about computerized voting?

The weirdest thing about the post-2000 movement towards computerized voting is that computerized voting, well, sucks. It's worth checking out the joint 2001 Caltech-MIT report on voting technology. Page 23 contains the following very striking table of "residual votes" (effectively uncounted + unmarked + overvoted ballots)

Machine Type Presidential Race Governor/Senatorial Races
Paper Ballot 1.8% 3.3%
Punch Card 2.5% 4.7%
Optical Scan 1.5% 3.5%
Lever Machine 1.5% 7.6%
Electronic (DRE) 2.3% 5.9%

DRE (Direct Recording Electronic) is the technical term for computerized voting machines. Note that optical scan machines (effectively scantron machines like the kind used for standardized tests) appear to be substantially more accurate than computerized DRE machines. Interestingly, the Caltech/MIT report attributes the residual voting rate of DRE machines specifically to bad user interfaces.

Why, then, the enthusiasm for computerized voting systems? The implied value proposition for these systems seems to be:

  1. They're easier to use.
  2. They're more accurate (possibly because they're easier to use).

However, as we've just seen, both of these assertions are extremely dubious. On the contrary, as far as we can tell, DRE systems are worse than more conventional optical scan systems.

I'm speculating here, but I think that the real motivation for computerized voting systems is something quite different. What prompted the widespread interest in these systems was the unedifying spectacle of the 2000 Florida recount, where election officials were forced to decide what the interpretation of individual ballots should be. Clearly, it's almost impossible to objectively set standards in such a politically charged environment.

What made the Florida situation possible was that the election officials had access to the raw data, namely the physical cards punched by the voters. The messiness of the raw data makes enforcing an objective standard very difficult. By contrast, DRE systems utterly eliminate the possibility of this kind of recount. The data is whatever the voting machine says it is. This has the advantage that we're not subjected to counting-time debates about the meaning of political votes, which is a good thing, since, as we saw in 2000, people's positions on the standard of measurement are almost inevitably contaminated by the result they are seeking. Perversely, the fix for this seems to be to destroy the data you would need to second-guess the result.

Of course, the problem with the current DRE systems seems to be that while objective, they're not very accurate. Worse yet, as Ed Felten has pointed out repeatedly, most of the current systems don't have any kind of audit trail, so detecting program errors and fraud becomes problematic. This doesn't seem like a very good set of tradeoffs. The question then becomes: can we design a system that is both objective and accurate?

Posted by ekr at 02:42 PM | Comments (12) | TrackBack

November 28, 2003

Some sample CD extras

Here are some extras that might appear on special edition albums:
Artist Album Extra
Led Zeppellin Led Zepellin IV 35 alternate takes of "Stairway to Heaven
Tupac Shakur Greatest Hits Documentary about Tupac's penal experience
Kraftwerk Autobahn 15 minutes of white noise
Yes Tales from Topographic Oceans Extended 37 minute remix of "The Revealing Science of God"
Motorhead Ace of Spades Lemmy explains the meaning of the song 'Love me like a reptile'

Just some suggestions...

Posted by ekr at 02:51 PM | Comments (52) | TrackBack

More ideas for the record company guys

The other day, I gladly forked over $30 at Borders for the Two Towers Extended Edition DVD set. Now, I'm never willing to pay $30 for a DVD--$20 is about my limit, no matter how much I like it-and I didn't even like the Fellowship of the Ring that much. So, why did I feel good about this? Well, there were four discs in the set, so I was clearly getting good value for my money. What those discs have is primarily a bunch of special features, documentaries, commentaries, etc. Surely when you add all that stuff up it's worth $30. At least, that's what I was thinking when I handed over the cash.

Anyway, this got me thinking about CDs. As EG readers know, I think that CDs are way overpriced (in the economic sense that they're more expensive than I'm willing to pay, even though I like music quite a bit) especially when compared to DVDs. The movie studios have gotten quite good at figuring out ways to get me to buy their product, and even to pay more than I normally would. Couldn't the music labels do the same?

Of course they could. Most of the stuff that appears on special edition DVDs is stuff that's sort of thrown off as part of the production process, or at least is very easy to make. It would be easy to deliver each album a double CD. The first CD would contain the album and the second one would contain the "extras". Things like:

  • Alternate tracks.
  • Interviews with the artists/commentary on songs.
  • "Documentaries" about the production of the album.

It's true that CDs sometimes contain extra tracks, but they're generally viewable only on your computer and I've never intentionally bought a CD with that feature over one without. By contrast, people absolutely are willing to pay more for DVDs with the special features, even if they never watch them--I'm having trouble believing I'll watch Two Towers 5 times (one for each commentary).

Seeing as the marginal cost of putting an extra disc in the package is close to zero and the DVD experience strongly suggests that people will pay rather more for the product if you do it, why don't record companies do this? I'm not sure. In general, the record companies don't seem that interested in giving consumers what they want--why did it take Apple to get us downloadable music--but the movie companies do. I'm not sure why.

Posted by ekr at 02:08 PM | Comments (29) | TrackBack

November 27, 2003

Why don't I get Advil XR?

I jacked up my back in the gym last Sunday and I'm on the standard 2400 mg of ibuprofen daily--at least that's the idea. Unfortunately, like many people, my compliance is less than optimal. I miss doses, and worse yet, I forget whether I took them or not. Either way, I get off my schedule. This got me thinking, though. Why do things have to be so complicated. Wouldn't it be a lot easier if I could just take one pill daily? Then I could take it in the morning and evening and just forget about it otherwise.

It's really common to make single daily dose drugs long-acting by putting a short acting drug in a delayed-release tablet. This approach is very popular with psychopharmaceuticals, e.g. Effexor XR or Adderall XR. But for some reason, there are lots of drugs which are short-acting but don't come in this kind of formulation. In particular, none of the over the counter non-steroidal anti-inflammatories (NSAIDs) do. Naproxen (Aleve) is long-acting, but as I understand it it's just that the drug has a long half-life, not that it's in an extended release formulation. Naproxen doesn't work as well for me as ibuprofen and I know other people who have experienced the same thing.

So, I don't understand why more drugs aren't available in XR formulations. Maybe someone who understands drug chemistry can explain it to me. Is there some technical problem that makes it impossible, or is it just not that the market is limited?

Posted by ekr at 01:28 PM | Comments (59) | TrackBack

November 26, 2003

Great, an RFID wallet

VeriChip, the guys who make implantable RFID tags, think it would be just a peachy idea to have an implantable RFID tag you could use to pay for stuff [*]. Now, the idea of using RFID tags to pay for stuff isn't new, and I'm not particularly weirded out at the idea of implanting stuff into my body, but I just don't see wanting this. The claimed rationale is that the chips can't be lost if they're implanted, but I'm having trouble getting too worried about that. I carry around lots of payment instruments every day (they're called "credit cards") and I know what to do if they're lost: I call the issuer and cancel them.

I guess the major benefit here is that you could use this form of payment when you're naked, which kind of limits the market to a few, uh, specialized markets.

Posted by ekr at 09:28 AM | Comments (13) | TrackBack

November 25, 2003

$200 off, huh?

Via Gizmodo, I see that Amazon is selling the Treo 600 with a $200 rebate on top of the Sprint rebate [*]. It doesn't look like it's shipping yet, though... Hmm.. I wonder if there's some way to get this rebate without returning my Treo and ordering a new one.
Posted by ekr at 02:18 PM | Comments (42) | TrackBack

November 24, 2003

Break before make

Ok. My Treo 600 is working again.

The problem with giving techies toys is that they like to play with them. In this case, I wanted to get PCS Vision Business connection working (it turns out to be totally useless for my purposes, BTW, but I didn't know that then.) Anyway, when I tried to get my Treo to talk to it, I kept getting a bad password error, so I decided to change my PCS Vision password (the wrong password, as it turns out, but I didn't figure that out till later either. In my defense, there are three different passwords associated with my service).

Changing the username and password is theoretically easy. You just change it on the web site and then the network pushes it out to your phone. So easy, in fact, that I decided to change my ugly automatically generated username while I was in there. This is where things started to go wrong. It takes from 2-4 hours for the network to update your phone and the network changes its password right away, so you can't use it until your phone is updated. Anyway, something went wrong; the update didn't work and I was off the net.

My first attempt to fix the problem was to just repush the password. No dice. The link had mysteriously disappeared. Then I called Sprint customer service, which did the same thing. It still didn't work. Finally, yesterday, I went into the Sprint store and waited in line for 20 minutes to talk to a cutomer support tech. They took it back to a tech and told me to come back in 40 minutes. When I got back, they said they'd reprovisioned it and it worked fine.

From a protocol perspective, this doesn't seem like exactly the way you want to design your system. If you've got a long, unreliable, update cycle, it's probably better to verify that the update actually worked before you deactivate the old account. This is especially true if fixing things up when they get wedged requires valuable customer service time.

Posted by ekr at 08:11 AM | Comments (13) | TrackBack

November 23, 2003

More on CAN-SPAM

As a security guy, here's the bit about the CAN-SPAM bill that I find curious. Under section 5(d)(1), senders of sexually-oriented commercial e-mail are required to put certain marks in the headers of their e-mail. These marks will be specified by the Commission but are explicitly intended to facilitate automatic filtering (S 5(d)(3)). By contrast, other commercial e-mails must contain "clear and conspicuous i entification that the message is an advertisement or solicitation" (S 5(a)(5)(A)(i)), but there's no requirement that the notice be machine processable. In fact, S 13(b) explicitly forbids the Commission from requiring any specific such marks:
LIMITATION.5 Subsection (a) may not be construed to authorize the Commission to establish a requirement pursuant to section 5(a)(5)(A) to inclu e any specific words, characters, marks, or labels in a commercial electronic mail message, or to include the identification required by section 5(a)(5)(A) in any particular part of such a mail message (such as the subject line or body).

So, you can automatically filter out sexually oriented commercial messages, but you have to actually view non-sexually oriented messages in order to filter them--after which, of course, you can opt-out. That's strange. If filtering works well for pornography, why wouldn't it work well for ordinary spam? This discrepancy makes me wonder if Congress is really serious about containing anything other than sexually-oriented spam. Is there some other explanation for this set of restrictions?

Posted by ekr at 09:36 AM | Comments (18) | TrackBack

Welcome to having an anti-spam law

You may not have noticed--I didn't--but the House has just passed an anti-spam bill and it looks like it's going top pass in the Senate [*], pseudo-cleverly called "CAN-SPAM" (The naming of bills is getting even worse than the code-naming of military operations. Enough with the cutesy-acronyms already. You're passing laws, not writing in to an advice columnist.). Anyway, John Gilmore is pretty upset about it:
This bill makes it a crime to use any false or misleading information in a domain name or email account application, and then send an email. That would make a large fraction of hotmail users instant criminals.

It also makes it a crime to remove or alter information in message headers in ways that would make it harder for a police officer to determine who had sent the email. Anonymizers will be illegal as soon as this bill becomes law.

There are MANY, MANY other things wrong with it -- including the fact that most of its provisions apply to *ALL* commercial email, not just BULK commercial email -- and that it takes zero account of the First Amendment, attempting to list what topics someone can validly send messages about, while outlawing all other topics that relate to commercial transactions.

If it passes, I think I can make a criminal out of just about any company. Companies are liable for spam that helps them, even if they had no part in sending it.

Read the bill yourself:
http://news.com.com/pdf/ne/2003/FINALSPAM.pdf
And weep. And then call your Congressman.

Everyone's common sense goes out the window when the topic is spam. They're willing to sacrifice whatever principles they have. And you already know how few principles Congress had left.

I've just given the bill a quick read, and here's what I took away:

  • It makes it illegal to anonymize commercial email. This sounds OK in principle, since it doesn't speak to non-commercial email, but anonymizers have no way of knowing if they're handling commercial mail, so I think it probably does make all anonymizers illegal.
  • It absolutely defines what's commercial or not and some of the restrictions only apply to commercial. It's not clear to me that this is actually a First Amendment violation. As I understand it, commercial speech can be regulated. I'm not saying this is a good idea, mind you.
  • I don't agree with John's assessment that you're liable if others promote your business. There's a clear exception in Sec 6(a)(3) if you take effort to prevent the transmission or detect the transmission and report it to the Commission. I imagine that this will be interpreted fairly loosely to avoid the kind of attack John proposes.
  • The bill requires markers for all commercial e-mail but the government is only going to prescribe markers for pornographic commercial.
  • Commercial email will be required to contain opt-out methods that senders will have to honor.

So, how much damage this bill does really comes down to how important you think anonymous e-mail is. I know John thinks it's really important. I'm not so sure. However, I do think it's somewhat important, so the question then becomes whether this bill will do any good or not. That's less clear.

The bill has a lot of introductory material but doesn't do that good a job of laying out the threat model. However, we can try to reverse engineer the threat model from the security measures the bill tries to apply. When we do that, we get something like this:

  • Most spam is sent by people inside the United States (because the bill primarily affects senders, it's not going to work by spam sent by foreigners).
  • People get multiple mails from the same spammer (because otherwise the opt-out feature won't work).
  • Spammers are people who aren't otherwise engaging in criminal behavior (otherwise making it double-plus illegal won't help much).

All of these propositions strike me as quite dubious, especially the third. A substantial--and as I understand it, increasing--fraction of spam is now sent through spam zombie networks. The people operating those networks are arguably already contracting for an illegal service--the penetration of the machines in the zombie network. Why would this new law deter such people? Moreover, I get a lot of spam that's pretty clearly sourced outside the US. This bill won't really affect those spammers at all.

Given all that, I'm not sure that this set of fixes is worth a law essentially prohibiting anonymous email.

Posted by ekr at 09:20 AM | Comments (73) | TrackBack

AskIgor

Here's something cool, via /. AskIgor is an automated debugging service. You give it your program and two input files, one that succeeds and one that fails. AskIgor tries to figure out what chain of events lead to the failure case (they call it a cause-effect chain) This looks like it might be a pretty useful debugging tool. I'll have to try it next time I run into a difficult bug.
Posted by ekr at 08:08 AM | Comments (8) | TrackBack

November 22, 2003

Upgrading my electronic brain

Terence and I went to the Sprint Store last night and I walked out with a Treo 600. I'm somewhat embarassed to report the the salesman pulled the old "we don't have any units left... wait, maybe we have one in the back" and I almost fell for it. However, since I'd walked in pretty much intending to buy the phone, I don't think it played much of a factor in my decision. Anyway, I'm pretty pleased with the unit.

So far, I've managed to:

  • Move all the data from my Clie onto the Treo. This worked flawlessly, btw.
  • Surf the web, including EG, which, btw, doesn't work on Terence's Hiptop.
  • Install Top Gun SSH (which, incidentally, doesn't seem to be that well integrated with the Treo's keyboard. Return doesn't do what's expected. But maybe there's some setting I need to do.)
  • Install VeriChat and do some instant messaging with AIM. This seems to work well, but I'm a little worried it's consuming my SMS quota... Need to check to see if I've got unlimited SMS.

Here's what doesn't work yet:

  • Jabber. I'm trying Chatopus but I keep getting disconnected.
  • I can't get logged into Sprint PCS Vision. It seems to be some kind of password problem and the web site says it may take up to 4 hours to update it.

I also need to get some MP3 playing software so I can use it to listen to tunes. Suggestions welcome.

Overall, I'm very impressed with the Treo so far. It seems like a really solid piece of hardware and I suspect I can get past the software problems. The keyboard is extremely usable and so far connectivity is good. It's a little larger and heavier than I would like, but certainly not unacceptably so. I'm definitely willing to pay the size price to get the keyboard. And because I only bought the 1 year contract, when the next model comes out I can switch over.

Posted by ekr at 11:28 AM | Comments (55) | TrackBack

November 21, 2003

Debian server compromise

The Debian project (one of the big Linux distributions) has just had a bunch of their machines compromised [*]. They're currently restoring from trusted backup. Ouch.
Posted by ekr at 07:27 AM | Comments (10) | TrackBack

November 20, 2003

Hello, Mr. FBI Agent, could you come fix my flat?

If you have a new car, it's reasonably likely that it has an on-board navigation and emergency assistance system. These systems are linked to the cell network and allow you to call for directions or help. Often, they'll call in automatically when the air bag deploys, so that you can get help when you're in an accident even if you're unable to manually call for help. The most well known of these systems is OnStar, but there's also TeleAid.

According to The Register [*], in 2001 the FBI obtained a court order to use this system for surveillance purposes. Effectively, it turned on the microphone in the car full-time and so that the FBI could listen in on the drivers. The operator of the system (the author of the article thinks it's TeleAid) sued to block the order and just won at the 9th Circuit Court of Appeals. However, the grounds for the appeal seem to be rather narrow: because the FBI was using the cell link full time, the car couldn't call for help in an emergency. So, it's entirely possible that a scheme that didn't interfere with the emergency service would be permissible.

It seems to me that this is especially bad news for the manufacturers of this kind of system. It's not common knowledge that you can use these systems in this way and now that I know it, I suspect that people who value their privacy will be a lot less likely to purchase them. Even if you're not worried about the FBI, there's no particular reason to believe that the system is resistant to random people activating it and listening in.

Posted by ekr at 11:40 AM | Comments (77) | TrackBack

November 19, 2003

Air time for the science guy

WARNING: the following posting contains an unusually high percentage of griping, even for me.

Tonight I caught the local public radio coverage of the Ebola vaccine trials. Now, the vaccine is an interesting topic and could have made a good story, since there's some pretty cool science going on. Instead they spent the entire story working the human interest angle about the the first subject in the clinical trials. Now, I sort of expect this kind of lousy, superficial coverage, but it's particularly annoying in this case since this guy is basically a peripheral player in the whole thing. He's just the first of a number of subjects in the trial. Would it be too much to ask to hear from the scientists who have spent years working on this project, rather than some guy who's major contribution is to be a human guinea pig.

Posted by ekr at 10:23 PM | Comments (39) | TrackBack

You have heard of quality assurance, right?

A lot of the concern about electronic voting has centered on the fact that the machines don't produce adequate audit trails and therefore are potentially susceptible to fraud. As Ed Felten points out, the problem is even worse than that. In some cases the software appears to have serious bugs that manifest under normal non-malicious use:
  • In Boone County Indiana, Microvote machines recorded 144,000 votes in an area where only 19,000 people were eligible to vote [*]
  • In Fairfax, VA, Advanced Voting Solutions machines appear to have selectively dropped votes for one candidate [*].

This, of course, is pretty much what everyone who understands software should have expected. Software has bugs, as anyone who has used a computer knows, and without really extensive auditing and quality assurance, you can't realistically hope to get even reasonable performance. Based on what I'm reading here and what I know about the Diebold case, the vendors don't seem to have followed even industry best practices for their code development, let alone the kind of careful programming techniques that really need to be employed for this kind of mission critical application. It's particularly troublesome that it seems to be multiple vendors who are having problems, since it suggests that noone is using good practices.

Seeing as the primary rationale for electronic voting is accuracy, what's needed here--at the very minimum--is a complete scrub of the code base. More importantly, as Ed points out, we really need human-verifiable audit trails of the entire voting process, since that's the only way to recover when errors are suspected.

Posted by ekr at 09:28 AM | Comments (9) | TrackBack

November 18, 2003

How not to choose code names

It probably sounded like an OK idea to name the crackdown on the Iraqi resistance "Iron Hammer", except that it turns out that Iron Hammer was the name of a planned German operation during WWII [*] (via Fark). Good choice, huh?

During WWII, the Germans used names chosen to be inspirational (e.g. the name for the invasion of the Soviet Union was "Barbarossa"). The US used to choose code names mostly randomly (with some censoring by the higher ups) but lately they've been chosen to be more inspirational as well. [*]. The problem, of course, is that there are a limited number of cool-sounding names and choosing good ones turns out to be tricky. This isn't the first time the US has chosen a bad code name, either:

After it declared war on terrorism, U.S. officials changed the code name for its impending attack on Afghanistan (news - web sites) to Operation Enduring Freedom.

The original name, Operation Infinite Justice, was jettisoned amid fears that the Muslim world, already leery of U.S. intentions, would object on the basis of Koranic teachings that only God can provide infinite justice.

Maybe we should just go back to random names.

Posted by ekr at 09:52 PM | Comments (32) | TrackBack

Would you like to get the Ebola vaccine?

Here's an interesting Washington Post article about trials of an Ebola vaccine. Apparently, they're having a lot of trouble getting people to volunteer to be vaccinated.
It is not surprising that people would hesitate about getting injected with anything labeled "Ebola." The disease is notoriously lethal, rapidly killing 50 percent to 90 percent of its victims. Just the thought makes people irrational.

Already, the landscaper [one of the subjects] has been asked by some clients whether he is going to be bringing the disease to the neighborhood.

Even though it sounds like the vaccine has been designed to be really safe (the article says "without a single component coming from the virus itself"), I'm not sure how eager I'd be to be one of the initial subjects either.

Posted by ekr at 08:32 PM | Comments (11) | TrackBack

November 17, 2003

Where do we get such salesmen?

As EG readers know, I've lately been considering the Treo 600. The key question I need to answer at this point is how well the roaming works, because, let's face it, Sprint's coverage sucks. Today, I set out to answer one simple question: Can you roam off Sprint's network?

First, Kevin and I went to the Sprint store in Palo Alto, where we received the rather implausible answer that you couldn't roam digitally but you could roam "using Verizon's analog network using TDMA". Seeing as TDMA is digital, but a different form of digital than what Sprint uses, that seemed like a rather implausible answer. Second, I called the Sprint hotline, where I was told that the Treo 600 was a single-band phone, despite the fact that Handspring's web site clearly states that it's dual-band (1900 MHz and 800 MHz). I think the actual answer to this question is that you can roam on Verizon's digital network (which is 800 MHz CDMA) but not on Verizon's analog network. However, it sure would be nice to hear the answer from Sprint.

It's hard to understand what the problem is here. Why can't I get a straight answer? Sprint's value proposition is that they have cool phones. Their major liability is that their coverage is lousy. Surely, assuring customers that they can roam should be a high priority, right? Apparently not.

Posted by ekr at 07:54 PM | Comments (54) | TrackBack

The future of the iPod?

Last night, Terence and I were talking about his iPod. Your typical album compresses down to 50 MB and since the big iPod has a 40 GB capacity, it can carry about 800 albums (about 700 hours of music). This kind of capacity is already pretty hard to manage. Just ripping all those CDs takes forever, as does downloading the data over the net. Even selecting all the albums is a big chore.

Here's my half-baked idea for how Apple can fix this problem and add additional consumer value: pre-populate the iPod with music. The way that this would work would be that you could buy the iPod pre-loaded with a large music collection. Because you're buying in bulk, you'd get a big price break, down to maybe $1/album. So, you could buy a bare iPod for $400 and a 75% populated one for $1000. Since the drive isn't (yet) large enough to store all the CDs in the world, I'd expect that Apple would offer a bunch of different packages. The easy way to do this is to offer just some primitive choices (Rock, Classical, etc.) [0] Kevin Dick suggested to me a clever scheme where you could buy "blocks" of music (e.g. the entire Barry Manilow collection) and then Apple would pre-load them for you. I'm not sure how easy this sort of customization is technically, but clearly Apple could send you pre-burnt CD images.

Why should the record companies go along with this? A number of reasons. First, it isolates them from risk. If people are buying in bulk without regard to the details of the collections that they're getting then the record companies don't have to worry so much about whether a particular album succeeds or fails. Second, it gives them an opportunity to extract more money. The record companies had revenues of about $12 billion in 2002. That's an average of only about $50/consumer. I'd certainly be willing to fork over way more than $50 to get all the albums for a given year on my iPod. Moreover, the record companies don't have to give away all their rights. The pre-loaded albums could be DRMed so that you couldn't pull them off the iPod. The customer would then be able to unlock them (maybe at $.25-$.50/song) and then it would be as if they had bought them off iTunes.

Finally, this scheme lets record companies expose people to new stuff. There's lots of albums I'm not willing to buy on spec, but if my iPod just put them on my playlist, I might find I liked the artist and buy the rest of their stuff.

This brings up the second important piece of the puzzle: better playlists. The iPod and iTunes is already a little smart about knowing what you would want to hear, but it could be a lot smarter, especially if it hard an enormous collection to choose from. It could occasionally throw on something new you might like--and after it had a while to learn your listening habits, would probably be right fairly often. And if it's not? No problem. Just hit the next button.

All of the technology is there to build this super-iPod. In fact, someone else has probably even suggested it before--though not to my knowledge. It's mostly a matter of building the right business arrangements. The popularity of iTunes and the iPod puts Apple in a fairly unique position in that respect. It would be nice to see them exploit it.

[0] This is also a great branding opportunity. For instance, you might be able to buy the Ozzy Ozbourne edition iPod, in black, with all of Ozzy's favorites pre-loaded.

Posted by ekr at 07:26 PM | Comments (55) | TrackBack

The Chechen National Anthem

Chris Bertram has a thread thread going on Crooked Timber about national anthems and what they represent. In that spirit, I present to you the Chechen National Anthem, found here.
We were born at night, when the she-wolf whelped.
In the morning, as lions howl, we were given our names.
In eagles' nests, our Mothers nursed us,
To tame a stallion, our Fathers taught us.

We were devoted to our Mothers, to people and the Native land,
And if they will need us - we'll respond courageously,
We grew up free, together with the mountain eagles,
Difficulties and obstacles we overcame with dignity.

Granite rocks will sooner fuse like lead,
Then we lose our Nobility in life and struggle.
The Earth will sooner be breached in boiling sun,
Then we appear before the world; losing our honor.

Never will we appear submissive before anyone,
Death or Freedom - we can choose only one way.
Our sisters cure our wounds by their songs,
The eyes of the beloved arouse us to the feat of arms.

If hunger gets us down - we'll gnaw the roots.
If thirst harasses us - we'll drink the grass dew.
We were born at night, when the she-wolf whelped.
God, Nation, and the Native land -
We devote ourselves only to their service.

I'm sure glad I'm not a Russian soldier in Chechnya.

Posted by ekr at 08:39 AM | Comments (12) | TrackBack

November 16, 2003

Singularity Sky

I've recently finished reading the quite interesting Singularity Sky. Singularity Sky is part of the recent crop of post- singularity science fiction. As with many such books, it primarily takes place among more-or-less standard humans, but with mostly unseen technological gods lurking in the background.

While not perfect, Singularity Sky had two features that endeared it to me. First, Stross really takes relativity seriously. It's hard to write galaxy-spanning science fiction without faster than light travel, but most authors assume the problem away using hyperspace or wormholes without thinking through the consequences. The problem is that relativity says that faster than light travel automatically implies time travel. [*]. Instead of ignoring this problem, Stross glories in it. It's one of the central plot points of the novel.

The second feature is a little more personal. Stross envisions a world in which the Internet Engineering Task Force is rather more important than it is today:

"But the UN is a government--"
"No, it isn't," Martin insisted. "It's a talking shop. Started out as a treaty organization, turned into a bureaucracy, then an escrow agent for various transnational trade and standards agreements. After the Singularity it was taken over by the Internet engineering task force"

Some future...

Posted by ekr at 03:21 PM | Comments (11) | TrackBack

November 15, 2003

Buying democracy

Gregg Easterbrook points out that [*] Dean and Bush have quite different funding strategies. Dean's average donation ($77) is less than a third of Bush's ($283). Easterbrook argues that this means Dean has to "sell out to everybody" and that this is bad:
So far Bush donations average $283, Dean donations average $77, George Will reports. Bush's average is high in part because the president's policies cater to the well-off. But having catered to the well-off, mainly on tax policy, Bush is relatively free to do as he wishes on other issues, assured that his campaign treasure chest will be full. This is, in campaign-finance terms, the One Big Sell-Out theory. Supposedly, Senator Robert Kerr of Oklahoma once said that the fact that he was totally sold out to and financed by the Kerr-McGee Corporation left him free to vote his conscience on everything else.

To draw lots and lots of small contributions, on the other hand, won't Dean have to promise everything to everybody? He hasn't yet. But as the initial glamour of his run inevitably wears off and the need to sustain cash flow begins to take over campaign strategy, watch to see if Dean begins endorsing anything that could inspire any donor group.

Now, it's not entirely clear from these averages that Bush actually is catering to a smaller group with more focused interests than that appealed to by Dean--the mean isn't enough to characterize the distribution--but let's say that that's the case. Why is it bad that Dean has to cater to more people?

Easterbrook's argument about Dean having to cater to a lot of donor groups seems totally backwards to me. In a democracy, politicians are supposed to be worried about the desires of their constituents. The reason that large political donations are distortionate is that they cause politicians to give greater preference to the views of donors than those of the rest of their constituents. The larger the donor base the more it will reflect the electorate--imagine if everyone who voted for Dean gave him equal amounts of money. So, isn't having more donors better, not worse?

Posted by ekr at 07:10 PM | Comments (60) | TrackBack

A hepatitis/asthma link?

Kevin Dick just pointed me to a really interesting article. There's been data lately that suggests that children who grow up in very clean conditions are more likely to develop allergies and asthma. However, the data so far was mostly epidemiological. Now it looks like some researchers may have found a molecular link between hepatitis and asthma--if you happen to get hepatitis A you may be less likely to get asthma. If the protective effect is real, the next step is to see if there's some way to get it without having to get hepatitis.

UPDATE:
I just noticed that this research was published in early October, but I just noticed it now...

Posted by ekr at 06:24 PM | Comments (11) | TrackBack

At least the Hilton understands security

The other day I went back to my hotel room to pick something up and housecleaning was in there cleaning it. Amazingly, the maid asked for my room key and then walked over to the door and--as far as I could tell--checked that it opened the lock. Nice work!
Posted by ekr at 06:09 PM | Comments (38) | TrackBack

November 14, 2003

I'm surrounded by Macs

At lunch today, out of 8 IETFers, 7 were Mac users and it looked like at least half had Mac laptops. This is slightly unusual in IETF, but a very significant fraction of IETFers have iBooks and PowerBooks. It's by far the highest percentage I've seen in any organization.

It looks to me like what's happening is that the Macs are replacing people's people's Linux and BSD boxes. Running free UNIXes on laptops has always been a pain since laptop hardware is often idiosyncratic and it can be difficult to get the OS configured, especially in the video and power management departments. Both Macs and Windows laptops come properly configured but since OS X is BSD underneath it's a congenial for people used to UNIX. Plus, you can run MS Office Apps. And since IETFers are much more likely to use free UNIXes they're more likely to switch to Mac.

I'm not quite ready to switch myself. The Mac UI just rubs me the wrong way and even the lightest Mac laptops are 50% heavier than my trusty Vaio. On the other hand, I've seen some pretty hard-core BSD-heads switch so maybe it's just a matter of time.

Posted by ekr at 10:14 PM | Comments (30) | TrackBack

Some data on Internet threats

Bernard Aboba's plenary presentation last night was... interesting. Notable points:
  • The volume of spam has been increasing very dramatically over the past year or so and there's no sign of it slowing down. (data courtesy of Xmission and Mark Handley)
  • The percentage of email that is spam/junk is increasing.
  • Viruses and worms seem to take a really long time to get wiped out. Virus probes on Vern Paxson's machines at LBL seem to be more or less constant. There was some theorizing that the entire pool of vulnerable machines has been infected.

No magic bullets were offered.

Posted by ekr at 06:53 AM | Comments (19) | TrackBack

November 13, 2003

This blog does not allow guns

At most of the Minneapolis establishments I've been in there are signs saying that this establishment doesn't allow guns. You never see this in California so I found it curious till Mark Crispin explained it to me. Apparently Minnesota has a "shall issue" [*] law that makes it easy to get concealed carry permits. However, such laws often allow businesses to forbid guns on their premises, and a lot of them have done so.
Posted by ekr at 06:15 PM | Comments (13) | TrackBack

Heard at IETF

"We're in a solution rich environment" -- Geoff Huston on the number of different anti-spam approaches.

"It says 'interference robustness' but it's more like 'interference fragility'" -- Vern Paxson on Mac OS X 802.11 settings.

"... or else the snake pit will go sliding down the slippery slope into the rathole" -- Paul Hoffman discussing a working group charter.

"Microsoft is so insular that people there actually don't realize that everyone in the world hates them" -- a recent Microsoft hire who hasn't yet drunk the Kool-Aid.

Posted by ekr at 09:02 AM | Comments (35) | TrackBack

A less pessimistic view of spam

Colby Cosh is arguing that we're about to win the war on spam [*]. His argument is that the countermeasures spammers have to use to get by Bayesian spam filters are so tortured that it's getting hard for humans to read the resulting messages. Cosh argues that relatively soon spam will be so cryptic that low response rates will make spamming unprofitable--in fact that it probably already is so.

I'm not sure I buy this argument. The spam I get is often kind of cryptic, but I have no real trouble reading it. Still, Cosh could be right it sure would be nice if he was.

Posted by ekr at 07:29 AM | Comments (63) | TrackBack

November 12, 2003

Futures markets

If you haven't seen the Foresight Exchange, you may want to check it out. It's an idea futures market that lets people bet on various future events. It's quite interesting to see what people think is plausible. For instance, people seem pretty bullish on the prospects for germ line gene therapy [*] and that Paul Krugman will win the Nobel Prize [*].
Posted by ekr at 07:16 PM | Comments (10) | TrackBack

November 11, 2003

The Treo 600 keyboard

So, I spent a while typing on Joe Hildebrand's Treo 600 for a while. It's true that the keyboard really is microscopic, but it's surprisingly easy to use. Without any practice, I was already substantially faster than Graffiti, and Joe says that after a day or two he's started to get better. Now, there's no Graffiti area, but apparently you can use Newpen and use the entire writing surface for Graffiti. I think I've found my new cell phone.

Now, the question is, "how tough is it to get your hands on?" Sprint claims that they're going to have number portability. I wonder if the salespeople know how to actually make it work.

Posted by ekr at 06:47 PM | Comments (11) | TrackBack

How heavy are your pants?

One of the guys here at IETF has a Treo 600. Nice little unit. I think we're now at the point where you can replace your PDA and phone with a single phone-sized device, thus reducing the load on your pants and/or belt. Now if I could just get a combination PDA/phone/leatherman...
Posted by ekr at 10:33 AM | Comments (36) | TrackBack

November 10, 2003

Not a good e-book but better than no e-book at all

Baen books has started publishing their titles online in advance of publication. In addition, in a fit of generosity they're giving away a fair portion of their backlist for free. I'm not a big fan of reading things online, but you can also download books to your Palm. Now, the Palm is a terrible e-book platform. You only can read about a paragraph at a time and the font isn't exactly attractive.

Still, it's a lot better than not having any e-books. I keep my Palm in a holster on my messenger bag so I can take it out and read for 5 minutes at a time wherever I am rather than digging through my bag for a book. And since I'd be carrying the Palm anyway, it's great to have 3 or 4 books on it with no weight penalty. Now if someone would just design a Palm-sized platform that had a good enough looking screen to really read on.

Posted by ekr at 05:58 AM | Comments (15) | TrackBack

November 09, 2003

Disposable cell phones

Via Slashdot comes a pointer to Hop-On, a vendor of disposable cell phones. I've seen prepaid cell phones before, but they're just cheap cellular phones with a prepaid calling plan. The Hop-On phones are clearly designed to be disposable from the get-go.

Disposable phones--and even old-style prepaid phones--really throw a wrench in the kind of wiretapping that the FBI has been so concerned with preserving. The other day when I was in 7/11 the clerk was perfectly happy to sell me a prepaid phone for cash without me showing any ID. I assume that these phones would be the same way. The FBI isn't allowed to monitor all cellular communications but only those from identified phones--or identified people--but none of that stuff works if people can get cheap and anonymous phones.

Posted by ekr at 09:47 AM | Comments (68) | TrackBack

Greetings from the land of static cling

I'm in Minneapolis for IETF58. So far I've noticed two very important facts about Minneapolis:
  1. It's cold.
  2. It's dry.

So dry, in fact, that there is a tremendous amount of static electricity associated with everything. After I turned off the lights, I could see sparks every time I moved under the sheets. I also kept shocking myself every time I got up, making going to the bathroom a terrifying experience.

Posted by ekr at 05:47 AM | Comments (54) | TrackBack

November 08, 2003

Shameless plug

For the past 2 years, Lisa has been working on a book on WebDAV. It's finally out. All proceeds from the sale of the book go directly into our pockets.

Seriously, though. Obviously I'm biased, but I think it's quite a nice piece of work. If you're interested in Web based authoring you should check it out.

Posted by ekr at 08:10 AM | Comments (12) | TrackBack

November 07, 2003

You don't mind if we hijack your web connection, do you?

It seems that Belkin has decided it woud be a good idea if their routers occasionally intercepted your HTTP connections and sent them to their own web site [*]. Belkin is apparently concerned that people who own their routers won't be aware of their parental control system, so every 8 hours they hijack one of your web connections and redirect it to a page advertising those services. Strangely, people find this annoying.

Belkin's defense, if you can call it that, is that once they've redirected you, you can turn off this "feature". Personally, I'd prefer if my router just, you know, routed, out of the box.

Posted by ekr at 06:54 PM | Comments (10) | TrackBack

November 06, 2003

Precious bodily fluids still in danger

In Tuesday's elections, the citizens of Palo Alto voted overwhelmingly to defeat Measure B, which would have removed fluoride from the drinking water. [*]. A good thing, too, as far as I'm concerned. The opposition to fluoride always struck me as, well, cracked.
Posted by ekr at 06:23 PM | Comments (11) | TrackBack

How should Terri Schiavo die?

I've been tracking The Terri Schiavo case [*] and I think it presents some interesting ethical issues. The background here is that Terri Schiavo had a stroke about 13 years ago and has been persistently vegetative ever since. Her husband wants to remove her feeding tube and let her die. Her parents don't want him to and after a lot of legal maneuvering (which the husband won) the Florida legislature passed a law letting Gov. Jeb Bush override the husband's choice.

Obviously, there's been a lot of argument about whether the feeding tube should be removed. I don't find it that interesting a question--though you might want to read Chris Rangel's excellent articles [*] [*] on this topic. But let's say for the sake of argument that she should be. Why bother with starving her to death? Why not just shoot her, or if that's too messy, give her an overdose of sedatives or something?

The usual distinction people like to draw here is between failing to help someone and actively killing them, (see, for instance, Kieran Healy's post here.) The idea is that letting something bad happen isn't the same as doing it yourself. But I don't think that's a very useful distinction in this case. The whole rationale here is that Terri Schiavo isn't really a person. Otherwise, it wouldn't be ok to starve her to death at all. Given that, starving her to death rather than killing her looks a lot less like a moral choice and a lot more like simple squeamishness

Posted by ekr at 05:13 PM | Comments (69) | TrackBack

November 05, 2003

What's the big deal with being root?

Modern operating systems like UNIX and the Windows NT series generally have a bunch of users with different privileges. Typically, you'll do most of your business as an ordinary user but there's an administrative user (called root on UNIX and Administrator on Windows) who is responsible for system administration tasks. So, for instance, ordinary users can't configure the network or add new users.

Obviously, an administrative user can do an enormous amount of to a computer, and so security types tend to pay a lot of attention to protecting the root user. So, for instance, there are lots of security holes which let an ordinary user escalate their privilege up to that of the root user.

Now, on a multiuser system, this makes a lot of sense, because you need to worry about user A accessing the files of user B. But most of these systems are only nominally multi-user. The Linux or Windows 2000 machines on people's desks are really single user machines. In the worst case, they have a bunch of services like web or mail servers running at low privilege, but lots of those machines are completely firewalled off.

In systems like this, it's barely worth protecting the root user. What's important on a single-user system isn't the system itself, it's your data and anyone who has your privileges has access to that. I'd much rather have someone trash my operating system--which I can reinstall in a day or so--then the data I've spent years accumulating. If someone breaks into your system through your mailer or web browser, then they've got the same privileges as you and you've got serious problems, even if they never get root at all.

Acknowledgement
A discussion with Hovav Shacham helped crystallize some of these thoughts in my mind the other night.

Posted by ekr at 11:05 PM | Comments (73) | TrackBack

Barbra Streisand on free speech

Barbra Streisand is complaining that CBS has cancelled "The Reagans" [*] (well, actually, moved it to Showtime). Now, I don't really have an opinion on whether the miniseries should have been cancelled. It sure sounds like a bunch of annoyed conservatives pressured CBS into cancelling the series.
Concluding her statement, Streisand launched into an American history lesson, stating the cancellation of "The Reagans" is "censorship, pure and simple."

"Well, maybe not all that pure," she continued. "Censorship never is. Due to their experience with the restrictive English government, the framers of our constitution specifically included a ban on prior restraint in the First Amendment, which is an attempt to stop information from getting out there before the public has a chance to see it at all exactly what is going on in this case."

CBS, as a private company, she acknowledged, "has the legal right to make decisions about what they do and do not air."

But she insisted "these important decisions should be based on artistic integrity rather than an attempt to appease a small group of vocal dissidents."

Of course, Barbra has every right to be annoyed about CBS's actions--just as the conservatives had every right to complain about the miniseries. I just find it rather ironic that she's simultaneously complaining that a private company decided not to show somethina and trying to use the power of the state to stop Ken Adelman from posting pictures of her house. [*] I guess the First Amendment only applies to speech you like.

Posted by ekr at 02:49 PM | Comments (53) | TrackBack

November 04, 2003

Another clever medical idea

The New Scientist is describing a clever idea for fighting cancer [*]. They create a bunch of gold/silica nanoshells attached to antibodies targeted to the tumor. They inject the particles and wait for them to bind to the tumor and then use IR light on the body. This heats up the nanoparticles killing the cancer. Clever, huh?

When monoclonal antibodies first came out, there was a lot of talk of using them to preferentially deliver poisons to cancer cells, but as far as I know not much ever came of it. Maybe this time will be different.

Posted by ekr at 08:37 PM | Comments (10) | TrackBack

November 03, 2003

Ipecac makes me want to puke

Terence Spies pointed me to this article about the American Academy of Pediatrics deprecating ipecac. The idea behind ipecac was that if you were poisoned your parents would feed you the ipecac which would make you vomit and expel the poison. Apparently, studies show that's not that great an idea
Tenenbein said poison control centers have been phasing out use of ipecac for a number of reasons. It sometimes causes prolonged vomiting and lethargy similar to drowsiness that might be caused by an overdose of sedative pills. Those symptoms can complicate diagnosis and treatment, Tenenbein said.

It also may not totally empty the stomach of poison, or may linger and cause a child to vomit up other antidotes, he said.

Another study in the same issue found that home use did not reduce emergency room visits and did not improve outcomes. It is based on an analysis of data from the American Association of Poison Control Centers, which also is preparing new ipecac guidelines.

I have a pretty clear memory of the bottle of ipecac on my parent's bathroom shelf, but as far as I know my family never used it. In fact, I think it's still there, 20 years later. Next time I'm at my parent's house I should check and see if it has fossilized.

Posted by ekr at 08:56 AM | Comments (10) | TrackBack

Making progress with organ transplants

Ok, this is kind of cool. [*]. Pfizer has a new anti-rejection drug for organ transplants. They designed the drug by inhibiting a protein that's broken in people with a specific immune deficiency ("bubble boy" disease). It seems to work in monkeys.

The background here is that organ transplantation is kind of a messy business. You see, your immune system doesn't like foreign tissue in your body and organs from other people definitely count as foreign tissue. When it detects the foreign tissue, the result is a powerful immune response called organ rejection which eventually results in the destruction of the organ. In order to stop this process, patients have take drugs to suppress the immune system. The standard drug, cyclosporine, pretty much ushered in the modern era of organ transplantation.

Obviously, this is dangerous and the trick is to suppress the rejection response without doing too much damage to the rest of the immune system as well. (Remember that what kills people with AIDS is that they don't have a functioning immune system). In addition, cyclospirine does damage to a lot of other cells in the body and is generally pretty hard on you. It looks like what the Pfizer researchers were looking for something that would suppress the immune system but have fewer other side effects. They did so by picking out and inhibiting a specific molecule that only is found in immune cells.

While this sounds pretty clever and is no doubt useful, it doesn't solve the real problem with organ transplantation, which isn't that transplants don't take often enough but rather that there aren't enough organs--not anywhere near enough. [*]. Even if we were able to get the conversion rate of potential donors to actual donors to 100%, there would still be a backlog for livers.

If we want to really meet the demand for organ replacement, we're going to need to be able to grow them rather than rely on cadaveric organ donors, which is why stem cells are so important. Hopefully, when we solve the stem cell problem, we'll also figure out how to grow organs that match up well enough that we don't need anti-rejection drugs.

Posted by ekr at 08:06 AM | Comments (59) | TrackBack

November 02, 2003

Comments spam

EG readers are generally a reasonably well-mannered bunch so I don't feel it necessary to edit the comments section. However, lately I've started getting spam in the comments section. In general, I do intend to delete that. Just so you know.
Posted by ekr at 10:25 AM | Comments (1) | TrackBack

November 01, 2003

Blacking out and deleting aren't the same thing

Say you're the government and you've got some embarassing document that's just been requested under the Freedom of Information Act. You want to blank out certain sections, but the document is in electronic form, so you do the electronic equivalent: you paint over those sections in black.

Unfortunately, as the Department of Justice just learned, this doesn't work well electronically. When you turn this file into PDF, at least in some cases, the output is the original file plus instructions to paint black squares over certain sections. These squares are trivial to remove and then you have the original document. [*].

This sort of thing actually happens fairly often. The important thing to realize is that just because you've deleted something in a file that doesn't mean that it's not lurking around there somewhere, open to anyone with a text editor. For instance, MS Word files often contain change histories for the file. If you just delete something and then publish the Word file, it's trivial to dig out the original.

The bottom line is that it's important to know your tools. Personally, when I have something condidential to send out, I print it out. That way I know exactly what the recipient is getting.