Bill Cheswick's Old Diaries

Start of Diary, at end of Lumeta time

Sep 20, 2006

When they are testing new diabetes drugs, do they use a sugar pill for a placebo?

Aug 13, 2006

I am attending a SCIFOO weekend at Google, given by O'Reilly. 200 scientists and a wiki conference. It's working great. Some notes:

Greg Bear is going on with his Darwin's radio stuff. I missed his book Darwin's children, and have to check it out. His most interesting comment concerns human pheramones: airport bathrooms are used by numerous alpha males, which is why we don't like to spend time in them. Well, there are other reasons, but it is an interesting speculation. Would the lady's room be more welcoming? (``Just running a scientific experiment, ma'am.'') He also speculates that non-productive matings still involve the exchange of retroviruses. Consiousness evolved as a social tool. When you are not thinking about how other people are reacting, you don't need it. His example: getting lost in your work, you literally lose your sense of self, perhaps a good sample of experiencing animal consiousness.

Robert Sawyer used us for ideas for some new book ideas. It's fun to come up with wild ideas, and let someone else do the work. I haven't read his stuff...gotta check some out.

GPHIN, a web monitoring system, spotted epidemic outbreaks well before WHO and countries did. Can this be improved with automation? (GPHIN involves seven experts scanning news stories on the web.) Catching disease outbreaks early is more important than getting just the right treatment. vicgdev.com is an attempt to take the experts out of this surveilance.

Vernor Vinge's here as well. He ran a session on EMP, my current major Badness scenario. We chatted a lot about writing, dialog, beginner mistakes, and things like that. I asked him if we would ever learn what's in the middle of the galaxy in the Deep books. He smiled. I am such a geek.

Following Verner Neil Gershenfeld of MIT talked about Internet0, which includes simple, cheap devices that send IP streams over a variety of transmission media, without relying on newfangled protocols underneath. These devices might be just the right price and performance for instrumenting houses, etc. I need to talk to doug.johnson@sun.com, who deals with distribution and commercial production of the microcontrollers.

Joshua Schacter pointed me to joshua/graph on http://del.icio.us, a web site I have to check out right now. His page has some interesting visualization software pointers I have to check out.

Coolest hardware: a camera that takes a picture, and reports the distance to each pixel within a cm. Works up to five miles, uses an IR laser that is eye-safe. From Stanford, I think.

July 23, 2006

I had an epic battle upgrading seismo to a new Intel mohterboard and CPU, now with enough umph to handle Vista. Gonna get me a beta...

The answers at the end of the session: You can build a bootable partition in FreeBSD is other than slice 1. Seems an odd thing to wonder about, but it turns out that the underlying problem is that the Fr4eeBSD install process can create a bootable partition, but creating an 'a' partition and populating it with dump/restore doesn't do the trick. There is something more, and I plan to ask someone rather than try to figure it out. Seems a shame: these quirks were less common in traditional Unix.

Also, either the boot ROM or FreeBSD doesn't mind clearing the "bootable" bit in the partition table, if the boot isn't going well for some reason. It took me a while to realize that this was happening, which was why good partitions were suddenly unbootable. I have to boot frequently from the CD/ROM just to set the bootable bit.

The Intel motherboard offered the usual array of confusing problems with disk locations and boot order. The new SATA drive (my first) could be any one of a number of drive numbers depending on BIOS settings, phase of the moon, etc.

The machine's up, the various clogged backup arteries are clearing, and I am moving a hell of a lot of data around just to clean up archival storage of stuff like personal data and the Internet Mapping Project.

Alas, I am way behind on important email, leaving good people hanging on various things.

I did catch Vint Cerf and Dave Farber on CSPAN tonight, debating Net Neutrality. They both had good points. In a deregulated world, without "natural monopolies", I don't think they would have had much to talk about on the subject. Heck, I had twice the network speed to my house two years before DSL was available than I do now. If congress feels the urge to do something, let's get this last (first) mile going. I still only have half a T1 to my home, and it costs way too much.

July 16, 2006

Myth confusion. I thought is was all working ok on the new 1.7GHz machine, including Hi-def. I came home to rampant stuttering on everything hi-def. Wassup? Did I blow the earlier tests? Did something change in the machine while I was gone? Are there new formats?

It does seem to be running near the edge, so I 1) overclocked the AMD chip from 13x133 to 14x133, giving me a 1.8GHz machine. I rebuilt myth to use XvMC, whatever the hell that is. Web pages said that the NVidia drivers don't handle it well, but it seems to work fine, and I am running 50% CPU usage on last night's high-def Saturday Night Live. With standard decoding, the CPU has only a few percent free CPU time. I wonder if I should unclock it. Now I can get back to a zillion other Myth things for the home, especially better music access.

CPU speed notes:

Detected 868.795 MHz processor.
6.5 set in boot
Calibrating delay using timer specific routine.. 1740.17 BogoMIPS (lpj=3480355)

default rom setting:
auto speed:
Detected 1737.451 MHz processor.
Calibrating delay using timer specific routine.. 3479.79 BogoMIPS (lpj=6959585)
13x

try 14x
Detected 1871.101 MHz processor.
Calibrating delay using timer specific routine.. 3747.40 BogoMIPS (lpj=7494808)

13 July 2006

I just heard a speaker refer to skinnydipping on the Internet. Looks like I've coined a new phrase.

15 June 2006

Project idea: The "host" portion of the IPv6 address space is an attractive area for hiding hosts in a large address space. There is an RFC with this idea in mind, and Timothy D. Morgan has a thesis proposing using this space for the prevention of spoofing. It is clear to me that there are a number of interesting approaches for using this space in ways not anticipated by the IPv6 planners.

It seems like that some may want to have a separate IPv6 address on their host for every TCP connection and UDP session. This will change the router's arp processing. The hosts on a network will see broadcast packets for everything, and have to decide for themselves if the connection is for them.

Here's the project idea: we need a mechanism in the kernel, and at user level, to implement wildly-changing IP addresses for a single host, in order to build test beds for these ideas.

23 May 2006

Kestrel cooked something tonight that contains "cummin". It smelled just like body odor to me, and I mentioned this to her. Apparently that description was way off for her, and clearly it smelled a lot better to her.

With over 400 receptors for smell, and a host of pseudogenes in humans for additional odors (totalling a noticable percentage of the human genome) it is clear that an assay is needed of the "primary odors", the chemical equivalent of the three (and sometimes four) primary colors in humans. I wonder if someone is working on such an assay. I guess it would involve perhaps a thousand or more compounds, some quite stinky, with the optimologists' question "does A smell like B"?

Though I just made up the term, it is no surprise that actual experts in the field have already coined the term, as google shows. The article in http://scifun.chem.wisc.edu/CHEMWEEK/Odors/CHEMOREC.html discusses anosmias, the inability to detect an odor. (Hmm, I wonder if there are sex-linked anosmias, similar to red-green colorblindness. The article says there are eight primary odor categories identified. The nose requires that volatile compounds disolve before they can be sensed, which would imply that a full-spectrum odor chip will have to be wet, and rinsed regularly.

13 May 2006

Today's assignment: think of some high-voltage power lines that run near your house or commute route. Use maps.google.com to find out where their connections are.

13 April 2006

> >In any case, don't get a meal ticket for me.  I seldom used them.
> 
> So what do you eat for 2 1/2 days?  Man can't survive on chocolate 
> alone.
> 
> Anne (But woman can.)
Part of my annual pilgrimage and calming trip to the west coast in mid-April is my 11am Wednesday visit to the Pacific Grove Safeway, where I acquire enough non-chocolate ches-food to last a couple of days. The collection of cheese, crackers, nuts, and fruits is often consumed in my vehicle while watching lovely ocean views during mealtimes. We New Jerseyites do love the shore, and the California coastline is quite exceptional.

I miss a lot of important conversations and the meals, but I am the pickiest eater you will probably ever meet (I eat like a three-year old) and standard fare is seldom appealing to me.

I am SO looking forward to this trip. Apparently, I need a break.

3 April 2006

I use one factor authentication in most of my daily work, except my laptop, which travels worldwide. There, the ssh keys are locked with a lengthy passphrase, giving me two-factor authentication.

It occurs to me that it is easy to add biometrics, a third factor, by measuring the typing statistics for that passphrase. It's processed locally, so there is no network jitter. It wouldn't have to be too strict. And the passphrase routine could be dropped in, leaving biometric information in a global variable for further processing. I wouldn't want the check to be too strict. But I bet I type distinctively enough that it might be an additional filter at virtually no cost. This wouldn't be a bad project for a grad student.

Of course, the whole thing could be a nice piece of indirection. It wouldn't be what you type, but how you type it. Spelling wouldn't count. Hmmm, I wonder if there are some typing statistics processing routines out there.

2 April 2006

Mythtv is mostly working again. But when installing the new frontend and bringing it up to 2.6.16-1.2069_FC4, the nvidia install shell is broken with the new kernel. conftest.sh incorrectly concludes that the kernel doesn't have vmap. I modified the conftest to return "4" for the vmap call.

I also installed patches from the net. I broke it up into three patches, p1, p2, and p3 because the full patch file was giving an error I didn't understand.

        make install
in src/nvidia. Do not use yum, or install any of the other stuff. The nvidia install seems to install the glx stuff as well.

the X system can be turned off with

        telinit 3
There are no further prompts, but X is gone. Then unload nvidia:
        rmmod nvidia
The nvidia works well, but this binary-only driver distribution is a constant pain in what should be a routine upgrade. I may get an ATI next time.

Myth is working nearly flawlessly now, except the backend crashes sometimes when starting up a new show. I suspect there are lingering drvier problems, though clearly the user program shouldn't crash. Given a little more free time, I will check out the crash. In the meantime, a cron job checks and starts a new backend every minute when there is none running.

24 March 2006

Opponents of GM crops are missing the point, and the boat. Mankind has been exercising genetic modification over plants, animals, and themselves ("you are not marrying that girl!") for tens of thousands of years. The mistakes of GM are likely to be minor, and self-correcting over time. The benefits are likely to be as useful as the work of Norman Borlang, possibly the greatest man in history, and you probably haven't heard of him. (It's been estimated that his work has saved a billion lives.)

These people ought to continue their protests and boycotts, but change the lettering on their signs and web pages to "boycott animals raised with systematic antibiotic treatments." OK, the words need improvement, but not the idea. The fact is, if you raise livestock on a steady diet of low-dose antibiotics, you get 10% more animal, which can be the difference between profit and loss when your competitors are doing the same thing. We want every one of these people to take the 10% hit across the boards because their antibiotic use and doses are idea for breeding antibiotic-resistant bacteria. They even use vancomycin, our current last line of defense against stubborn strains.

I'd gladly pay a 10% tax on meat to increase the likelihood that the next potent antibiotic I need is more likely to work.

24 March 2006

Saw Spike Lee's "Inside Job" last night. Excellent movie, and plenty to think about afterwards.

23 March 2006

I've decided to write up my old algorithmic password project and associated ideas for the Hot Topics co-conference with Usenix Security in Vancouver. Working title: Johnny can hash, a little. It's half-baked, but authentication is important and the world is seeking some new ideas. Anybody know a major league baseball coach (possibly retired) who would be willing to give me an hour or two of his time?

22 March 2006

I've been reading about a jillion papers in conjunction with my program committee duties. One of the best is the HoneyMonkey paper from Microsoft. I'd like to see more pro-activity like this.

5 March 2006

Mythtv. Got back from a 9 day trip to Europa to settle down and watch the Olympic. All that figure skating in HDTV, plus hours of the finals of curling.

Nope. The HDTV stuff was hopelessly infested with dropouts---unwatchable. The off-channel curling stuff was Ok. I am spending another weekend minutes away from a working system, it seems. Is DVB broken in the latest Fedora Core 4? Why does ivtv mess everything up? Why did all of this work before? How can Linux be so popular when it seems to be so damn brittle? And perhaps I have a hardware problem. Gosh, I think I know a thing or two about system administration...

5 March 2006

I set out a bunch of old-fashioned mouse traps yesterday. There has been at least one mouse in the attic waking us up for months, so it is about time. Usually, I perform the trapping/killing ritual in the fall. Dead mice are laid out on the lawn, belly up, and are soon gone. It's a bird feeder.

We have used the traditional mouse traps, and the "Have-a-Heart" traps of various kinds. In fact, https://cheswick.com/ches/mouse/index.html has a little non-snuff film I made with an infrared camera.

It seemed to me that the have-a-hearts were a bit misnamed. I am sure that some people check their have-a-hearts regularly, and release the captured rodents some distance from home. I did that once, releasing a mouse in the middle of a large, sunny, empty parking lot. The mouse got the hell out of there before a passing bird got a snack. But if you forget to check the trap, the contents tend to die very quickly. They have high metabolisms and quickly die of dehydration. I figured the old-fashioned kind did it almost instantly: SNAP, broken spine, gone.

So I bought some traditional mouse traps at the local Shop Rite, a bunch of them. We might have a lot of mice. (Statistics quiz: estimate the number of remaining mice based on the capture rate. Assume Poisson distribution of mouse arrivals. I just saw a related problem in a paper I read this week. Shades of a math class from 30 years ago.)

Anyway, the Shop Rite mouse traps are un-settable. I can't do it, without serious risk to my fingers. They are going back. I bought even more traditional traps yesterday at the hardware store. ("Everyone goes to the hardware store." -- Click and Clack) The hardware store had a $30 electrical mouse trap, guaranteed to kill them in seconds. Safety interlock. Needs 4 AA cells.

I set the traps all over the house, and checked through the day. Nothing. Then, at first light this morning, sudden loud squeaking (well, SKF (a cat) and I heard it), and it persisted for about a minute. It was upsetting. I have no particular need to cause my fellow mammals intense pain; I don't need to drive a tiny wooden stake through their little hearts. And I would rather go that way than die of thirst. Maybe the electrocution trap would be nicer.

At least a bird will have a little something to eat.

4 March 2006

If you take two laptops on your carry-on luggage, you have twice the fun at airport security.

2 March 2006

Our national energy policy should not be using research dollars for alternative energy: that will come as it does, and we have wasted enough money on such approaches. The stated goal should be "cheap electricity", to be met by a passle of cheap, safe nuclear plants. Regulations should emphasize safety, especially in the face of attacks, but do so reasonably, not with punitive regulations designed to destroy the industry. Plants should be of a size that they can be safely disassembled and removed if the need arises. I suspect fluidized bed technology would be about right.

If electricity actually gets cheaper, and relies more on nuclear technology, we can start getting out of the carbon game, which is currently very cheap and also politically unpopular. Very cheap electricity implies electric heat for homes, removing some natural gas and oil consumption. It provides a source of energy for automobiles, which really are stuck on petroleum for a while. Cheap electricity would improve the economics for batteries and even hydrogen in the automobile. Also, cheaper energy raises all boats, helping industrial and economic efficiency.

2 March 2006

Sci fi idea: aliens come and live in earth. They are tall. They learn the concept of graffiti. Now our bridges have six-foot high regular graffiti, and also a much higher layer scrawled by the aliens.

1 March 2006

Presenting the Elgin Powerpoints.

27 Feb 2006

We have encountered another B&B with RSSS, Remarkably Stingy Scottish Shower. This is an "electric shower", and appears to provide approximately one litre/min of hot water. (I actually considered measuring the flow rate.) It has a high, medium, and low volume setting, and I am hard-pressed to imagine someone using anything other than the high setting. There is also a "low flow" warning light, which presumably comes on if there is insufficient flow for the heater. That would be a remarkably low volume. This doesn't necessarily appear to be a money-saving gadget, as it was mounted over a traditional bath tub with two taps and lusty flow of both cold and hot water.

On the plus side of UK technology, round-abouts continue to work just fine. I think putting one in at the intersection of Mine Mount and Claremont in Bernardsville would solve that intersection's problems nicely. It's not quite a direct meeting of the five roads, so some curb rearrangements would be needed, and, of course, five yield signs and five rotary signs to teach the locals how to drive it.

It was good to visit Kestrel in Scotland again, but I don't need to do it for a while. I've seen enough of Edinburgh, and she is busy with school and a social life. We are on to Hamburg, our first visit there since the Danes won the World Cup. The joy we encountered in Denmark on that trip was matched this time by the Scots rejoicing over their Six Countries rugby victory over England. We saw the last part of this game and, though we don't quite understand all of rugby's rules, it was quite exciting. The last play of the game ended with a player kicking the ball into the stands, which I presumed ended the play with the time left well into the negative.

21 Feb 2006

I used to play with liquid nitrogen a lot when I worked at the Labs. The stuff costs a few cents per quart, and it was free at the Labs. Now I want to try some of this: http://www.boingboing.net/2006/02/20/fun_with_liquid_nitr.html

21 Feb 2006

Forget nukes. They are nasty, an all, and people are going to misuse them, and probably pretty soon. But that's the hard way. Here are my concerns, as written in the Foreword of Steve Brannigan's book, and here is an example of how easy it is going to be.

14 Feb 2006

Sci fi idea fragment: In near-future world, bad guys build a nasty but non-deadly virus. Their motive is profit through extortion. They announce the coming virus, and prove it is a problem by releasing a much tamer, unrelated virus with distinct symptoms. They then advertise a vaccine or antidote, distributed through underworld channels, for a price. They release the virus. People are very angry with them, and law enforcement tracks them down. In the ensuing shootout, all the guys with the technical knowledge of the antidote are killed. They leave behind an encrypted disk, without the key. (This is all back story to the actual story, which I have no ideas about.) Now, most of the world is running a dictionary-attack screen saver to try to crack the antidote (or vaccine), while the plague continues in an ongoing infection. Great chance to talk about the motivations of current virus writers, the coming dangerous competence with manufactured life forms, the future of computing and the Internet. Also, the variations and moderation of virulence in a long-running plague, and the associated host/pathogen evolution. Perhaps a nod towards Paul Ewald and the evolutionary biologists, who are rock stars in this new world. Also can deal with issues of quarantine and the purpose of government, and perhaps a logic for a world government, or at least police force.

14 Feb 2006

I am reading a lot of papers about darknets, packet telescopes, and reactive honeynets. I wonder if it is time for a packet telescope telescope. Perhaps even an Internet-wide audit at the class C address space level, or so. If done slowly, and sticking with routable addressing, a careful sampling could probably be done without annoying the denizens.

14 Feb 2006

I've just checked out "metasploit", something that has been on my list of things to check out for a while. Their documentation needs a justification section. Calling it a "framework for exploit research" isn't very specific.

14 Feb 2006

I've noticed that cats don't put things back when they are done playing with them.

11 Feb 2006

Mythtv is excellent for viewing the Olympics here in the states. The skip-forward feature makes quick work of the barrage of commercials, and the soft, non-Olympic up-close-and-personal bits they throw in. The scheduling works well, too. It knows about the repeated segments and doesn't record them, which is nice. The second, non-HDTV card is catching the hand-entered alternative "minor" sports, so I expect to get all the curling I can stand. And the HDTV is gorgeous. The opening intro to the human sacrifices on Friday night had postcard-beautiful images of Torino.

Of course, the alternate card isn't working in record mode at the moment. At least I got the sound back. I need to figure this one out before 3 AM Monday morning, when the first curling comes on. I am skipping the hockey games, at least until the semi-finals.

Back to current diary.

Back to my home page.