Bill Cheswick's Diary

15 April 2007

I always thought there might be a useful service, or even a business, around backscatter. Support Intelligence seems to have a good one.

Rimonabant Seen Emerging as the Top Diet Drug in 2010.

Apparently, my web "TV" appearances are, or will be at, here. I don't run macromedia on this secure mail-reading host, so I haven't checked it out. I have been running into people in Florham Park who attended. Apparently the audience doesn't come just from the Middletown AT&T site.

How I summarized the sabbatical:

Ches is now recovering from a six month (voluntary) sabbatical, 195 consecutive Saturdays. Sabbatical activities included a visit to the Supreme Court to hear oral arguments on KSR v. Teleflex (what should "obvious" mean in patent law?), a couple appearances on Hugh Thompson's new web "TV" show, program committee work, science museum software, some family visits, and not cleaning out the garage. He's rejoined the official research world at AT&T, which shows a lot of promise under its new owners. He's not sure what he is going to work on next.

"Retirement" a.k.a. sabbatical

3 April 2007

Memo to selves: read up a bit more on "concreteness" in the Kucera-Frances word frequency list (1967).

29 Mar 2007

I am filling out forms, and going through legal documentation, to start work at AT&T Research on the 9th. Very few Saturdays left, but I am psyched. This sabbatical has been helpful in a number of ways, which I should write down.

Drew Gross points me to a nice visualization of papers and their references. The graph isn't that large, though. I continue to wonder how to visualize giant balls of yarn in graphics, and have been working with Tamara Munzner and a student of hers, Daniel Archambault, at UBC, who is taking a crack at the problem. This might be something I want to work on at AT&T.

28 Mar 2007

On the day Al Gore spoke to Congress, Bjorn Lomborg did too. The media ignored the more useful of the two testimonies. But CSPAN has it, and it is easily found by searching for "bjorn". Alas, it is available via RTSP. If you are behind a NAT device, rtsp is not available to you unless you open an array ports (bad idea, but RealAudio usually fought firewall-friendliness) or run a proxy.

There is a reference implementation of an appropriate proxy, osrtspproxy, in the ports collection. Alas, it needs modification to be run reasonably on a gateway machine. It listens to a port (TCP 554), but listens on ANY interface, which means a dual-homed machine offers the useful end of the proxy to outsiders: bad idea.

The realplayer documentation also says that socks will work, and it ought to. I am using Windows Media Player (ver 11) for this exercise, because in my experience realplayer tends to come with a sea of noxious spyware-like programs I don't like. I wonder if mediaplayer can use, it doesn't. I will have to run the rtsp proxy, and just tear it down when I am done.

25 Mar 2007

Funniest comic strip I have read in months, today's Pearls Before Swine.

20 Mar 2007

The colorblind science exhibit is running nicely, though the layout could be more professional, and it could use more images, such as flowers and the like. Also, I'd like to include several of those colorblindness test cards.

Imagine how many tests one will need for the various forms of hard-of-smelling!

19 Mar 2007

Dan Klein has friends with parties similar to those we through in Bernardsville from time to time. Apparently, someone thinks this is commercial activity. Sounds overbearing to me! for details.

I've got a consulting gig. Fascinating, but I can't talk about it.

17 Mar 2007

I got the colorblind museum exhibit working. Need to get pointers up on the web page. The Children's Discovery Museum folks are complaining that their ancient hardware is dying and they need their new software. This leads me to the FreeBSD livecd issue: it would be nice to provide ISO images for each of these exhibits, including all the source code, for easy distribution and testing.

Livecds are clearly important, but there are too many choices. Picobsd is deprecated. Nanobsd is for memory stick disks, and seems about right, but I want to use CDs. LiveCD is a lot of work. I got it working about three years ago. Who knows how much has changed. Tinybsd seems nice, but doesn't seem to work off a CD-ROM. Perhaps FreeSBIE...

This shouldn't be a hard problem, I think. I want to create a FreeBSD Live CD that autostarts a program I supply in a directory I supply. The minimum call might be:

	build_livecd mydir mydir/start
dynamic libraries should Just Work. The CD should contain my current kernel (perhaps rebuilt: KERN=LOCAL would be an acceptible parameter) and installed ports. This might actually be hard to do, alas.

27 Feb 2007

Mytharchive. Finally got this working, so I can send some shows to my daughter in the UK, where House doesn't seem to be available for non-TV people. The notes:

I brought the mythtv-0.20 tree right up to date, recompiled and reinstalled everything.

There was a problem: one of the programs couldn't find dlclose. A quite googling showed that I needed these three rpms from atrpms, not freshrpms:

Now it's working, though the slowness is astonishing. To burn one DVD with a single episode of House seems to take about five hours of various scans and recoding on a 1.8GHz machine. The DVD works fine in our standard DVD player.

25 Feb 2007

< a href="">This is not a bad start at improving security: a bullet list of security features within an operating system. It is missing a number of features that I think are more important, like sandboxing technologies available, but it is a good start.

27 Jan 2007

Jefferson's wife upon discovering $90,000 in the freezer: "Dear, we are out of ice cream, but it is totally not a problem." -- Jane Condon

Excelsior, you fathead! Fred Avolio pointed out that a great deal of Jean Shepherd's broadcasts are available on the net. I wrote a script to gather copies. Shep would have approved, I think. I now have plenty for exercising with the Nano.

23 Jan 2007

As near as I can tell, I was slashdotted today. It probably had something to do with a by-email interview I have been working on for four months. It came out today.

In any case, my little 786Kb DSL line to my larger Dell laptop server for is quite busy. We are going to the library to see if we can read stuff from there.

17 Jan 2007

A nice visualization of different voting systems:

3 Jan 2007

Finally learned how to spell sabbatical today.

Beethoven, at last

Lorette got me an Ipod Nano for Christmas, something I wasn't sure I wanted. I've loaded a few things in, mostly Beethoven, to try it out. (Its capacity is 2GB.) It's a wonderful form factor for some situations, and I think I am going to use it all the time while travelling, or working in a noisy environment.

I am listening to Beethoven's 7th (von Karajan, Berlin). It's fast, of course, and glorious. I shouldn't have let it go so long, and Lorette and I are planning a few trips into NYC to get better fidelity, though these silly Apple ear plugs are not bad.

Sadly, I used to listen to these in my youth with my Dad. This is the first listening since his death a year ago. The moment's not teary, but definitely bittersweet. He once bought tickets for the whole family to hear the 9th at Avery Fisher with Bernstein (a personal favorite) in the center of the hall, ten rows back: his favorite place.

3 Jan 2007

Obesity Drugs

25 Dec 2006

Got an Ipod Nano today. Didn't imagine I would have this much fun with it. But here's the deal. I went to to upload itunes, and it seems to be very busy.

Of course it is: all those new Ipods are getting opened and loaded today. How many? I don't know, and we probably won't until the next Apple earnings announcement. But web activity might contain investment data, and web sites can be sampled for activity, both in response time, and by monitoring connections with the ip_id value, in many cases.

I wonder how many people do this. I can't have been the first to think of it. I wonder how strong the correlation is.

13 Dec 2006

My email inbox is empty. This only happens occasionally, maybe twice a year. What kind of stuff lingers, and is hardest to get through? Messages containing or pointing to papers I want to read. Queries I haven't figured out how to answer. And things that require Windows to view, usually videos of some sort.

This was one of the goals of this sabbatical. I will attempt to keep it empty, just as I am keeping the newly-cleared HSCs (horizontal shit collectors, a.k.a. desks) cleared.

Projects high on the list:

  1. Update our phone system to Vonage/Skype/something, cut our phone bills, and rewire our cut-down blocks to clean up the wiring mess and make it easy to change and configure this stuff.
  2. Update the Liberty Science Center software to answer their current requests.
  3. Add buy/sell markers to our stock charts to give a visualization of the context of our buy/sell decisions.

7 Dec 2006

Scut-work Thursday. Going to get the email list down to zero today. Twenty five to go.

Bussard and electrostatic fusion

Why isn't someone funding this guy, or chasing down his results? It reads like old-time sci fi, and includes a payoff of the scale of Heinlein's Harriman. For way under a megabuck, aneutronic fusion seems darn feasible, and not with the giant super-conducting catherdrals like ITER.

There are a number of clear technology wins, past and pending, that ought to be obvious. Napolean had a set of aluminum silverwear that cost more than solid gold sets. With aluminum as common as it is, this was clearly going to change, eventually.

Diamonds are just carbon, with decent heat content. It is only a matter of time before people will be able to afford to burn buckets of flawless blue-white diamonds in their fireplace.

Energy shortage? We are soaked with it, from the sun, from fossil (or maybe not: Gold's hypothesis? We really don't know.) fuels, or from matter. There is plenty of energy, and we need it. Recently Paul Harvey decried the cost in hot water of singing in the shower. This parsimonious view of the future, and of human potential, is more typical of hand-wringing liberals. Our goal should be that the poorest peasant in the world should be easily able to afford the longest hot shower he wants, without destroying the planet.

Bussard's talk is at

Home directory, or live off laptop?

If you have your home files on a laptop, you have to carry the laptop everywhere, even on your daily commute, or leave it hooked to the web somehow. Another plan is to have a home machine or directory and either connect to it, or import its file system. Another approach is to put all your files in a CVS or svn tree and update and commit from various sources.

I have done all of the above over the Lumeta years, and it has worked fairly well, but having one directory tree to bind them is really the way to go. So I am using NFS for the first time---it isn't much of a threat to the client, and my server is not directly reachable from the Internet. I wish I could use NFS v4, but the development money ran out at CITI and FreeBSD doesn't support it. I could switch to Linux, but Mythtv has taught me that it is too brittle. At least FC is.

I will have to tunnel remote NFS connections, probably using TCP calls over ssh tunnels.

SCOTUS notes

It was awesome to attend the oral arguments last week, and I don't use that word lightly. The building is nice, and the security is tight, but the awesome part is that j-random citizens (or even those like Lorette and me with a VIP pass) can watch the highest court at work.

Our connection through Steven Breyer got us on the VIP list, so we skipped the line (well, the external line) and entered along with the many lawyers.

We got the front row of the gallery, at 9:30. We got to sit quietly for 30 minutes. You are allowed nothing but pad and paper, and women are allowed purses, but absolutely no electronics are allowed. Lorette's sunglasses that hung around her neck were not allowed, something about reflected light bothering the justices! (I am guessing some justice complained in 1922, and the rule was laid down.)

At 10am a buzzer sounds, we all stand, someone does the "Oi yey Oi yey speech (maybe from the Talmud? 8-)), the nine enter, and we all sit down.

About two dozen lawyers were offered up and accepted before the Supreme Court bar. A group of lawyers from Union county NJ all seemed to have pointy noses.

There were two cases: an anti-trust problem with Weyerhauser and KSR v. Teleflex. Though we knew nothing about the first case, it is still fascinating to watch the justices and lawyers interact. It's also fun to compare the stereotypes of each justice with their interactions. There were no surprises, and a number of laughs. I have now attended 2 hours and 50 minutes of oral arguments, and I have never seen J. Thomas say a word out loud. But he is wide awake, whispering to Breyer next to him, and having one of the acolytes behind him pass notes to Scalia.

It is a challenge trying to figure out how the justices are going to rule on a case based on their oral arguments. The tapes and transcripts are online, so you can listen and judge for yourself. (I've actually been Mything the Saturday afternoon CSPAN "America and the Courts" to follow some of this stuff.)

If I had to guess, I'd say that the obviousness test at least in patent cases is going to change, and maybe a lot. But the justices are quite concerned that overturning a 1982 Circuit court standard will precipitate 100,000 patent lawsuits.

I look forward to reading the decisions.

(Busy day, scutwork Thursday)

I've started centralizing my files, and using NFS links to bring my web pages into my centralized directory. I found something I wrote early in the year that hadn't made it to the web pages: a dream I had on How to Trisect an Angle with a Chainsaw, by Arjen Lenstra.

4 Dec 2006

Broadband to the Home

We are paying rediculous prices for broadband to our home. So, finally, I have called some broadband providers.

At present we are paying $160/mo for half T1-speeds over a DSL line with 16 static IPs (13 usable, of course) provided by, who has served us well over the years. But this rate and speed has hung around for years, and it hasn't been getting any better.

FIOS is getting hung in B'ville right now. We've actually talked to the lineman. I checked their service. Residential service explicitly rejects running any sort of server at home. (This is short-sighted on their part: I see a lot of future demand and business for servers at home.) I could cheat in an number of ways, perhaps by spoofing or tunneling packets and keeping my current static IPs at a lower speed.

But the FIOS for business, though a little pricey, is still much cheaper, and gives me what I want. For $100 we get 15/2Mb speeds and 1-5 static IP addresses for no extra charge. There are also further deals with phone and video, but I suspect that we will run our own phone stuff over the fiber. It's time to check into asterisk.

Slides for Presentations

One of the things I do best is give talks, and I like to run through a lot of slides. My rate is perhaps 0.10 Lessigs, which is pretty darn fast (he does 800/hour) and disconcerting especially to ESL people. I have been using Powerpoint for 10 years, and though it has its moments, I want slides that are: One solution is to wait until I get my Mac, which is real soon now. (Listening, Santa? If not, I may have to replace you with a Visa card or (*Gasp*) an employer!)) It seems to have a nice slide program, though I have no idea if it meets all the criteria.

Another solution, one that I have been toying with, is to use one of the LaTeX slide packages going around. I have tried a few: prosper, beamer, and SMB-recommended powerdot. The latter appears to call latex/dvipdf for each slide, making slide generation very slow. I am used to generating entire books in a few seconds on modern CPUs.

Here is a rough list of the features I want:

  1. not much overhead per slide.
  2. bullet items and simple text available. It would be nice if unviewed bullet items where light grey, and revealed as used.
  3. on-slide Bibtex entries: references appear at the bottom of the slides on-which they are cited
  4. slides-left display on the bottom: slide 3 of 82.
  5. The sections are gathered into a table of contents, which is the introductory slide
  6. A slide with the Abstract, I think. Nice to keep the abstract with the talk.
  7. Not too awful to include pictures.
For item 1, I am probably stuck with:
which is ok, I guess. Then there is the embedded \begin{itemize} \item \item ... \end{itemize} which makes for a fair amount of boilerplate for each slide. This also adds clutter to the source code of the presentation. I might end up creating a little language that runs through m4(1). It would look a bit like troff!

The slide package that seems to roughly fit the bill is beamer. There's an awful lot there, way too much. And it has font problems on my computer. I'll work on this, along with updating Unix on My Mind talk, this afternoon.

29 Nov 2006

I've found a new use for google maps: assessing the size of Christmas tree farms.

20 Nov 2006

Oops. It has been a while. Haven't talked about the Supreme Court oral arguments (so cool we are going back next tuesday) or Hackers. But for today:

Web site idea

Making boring things interesting, through time lapse. I am not sure if paint drying would be more interesting sped up, but grass growing certainly might. What is is boring? Google isn't as immediately helpful as I'd like: those two are the quintessential boring activities. This Dull Men's Club has some amusing suggestions.


For years I have referred to an old program, crashme, that killed a surprising number of operating systems by making random system calls with random arguments. This idea was extended to sending packets with correct checksums but random headers, contents, etc. I was able to crash some TCP/IP implementations doing this in the early 90s.

This site uses several actual tools to do this stuff now, programs list sysfuzz, fsfuzzer, etc. These, along with the latest Minix test programs, would not be a bad start for a large collection of testers and verifiers of system calls.

And regression tests we need, c.f. Setuid Demystified and lots of more recent papers. I think these efforts offer a fruitful line of work for current researchers, and will help make the Internet world safer.

28 Oct 2006

This is a brilliant talk showing clever visualizations of world economic, health, and other data over time. If you are worried about the gap between rich and poor, your information is incomplete if you are not familiar with this.

Woke up this morning with some ideas for a talk on computing and security futures. Includes the "hollywood shell", which I think I'll write today.

25 Oct 2006

Trying out the new Dell 2407FP LCD. Very nice. The modeline:
    Modeline "1920x1200"  154.0   1920 1968 2000 2080  1200 1203 1209 1235

17 Oct 2006

Myth, final installment for a while

Here are more answers:
  • 1.7Ghz - 1.8GHz CPU just barely drives an NVIDIA card at max hi-def without losing frames.
  • With brand X chips, 2.4MHz is not enough.
  • 64-bit Athalon 3400 (2.4GHz, thanks guys) with NVIDIA 5200 chip has 30% idle CPU with kernel deinterlacing and libmpeg2
  • something appears to have happened between April and now to make Myth less efficient, which is why I had to switch out the installation that had been running fine before.
  • Even with myth installs (at least for the front end machine), it is much better to install FC5 cleanly on a machine, and then do a clean myth install following Jared's instructions
  • NVIDIA and lirc yum installs Just Work. Don't need to visit NVIDIA for their stuff. Thanks!
  • Perhaps Myth is actually getting a little less fragile, and a little easier to install and deal with.

    Now I can clean up all the PC shards lying around, and tuck the new machine away where the fan won't be instrusive.

    Thinking of going bowling for my daily exercist today. It is quite a workout if you use a lane of your own for an hour.

    14 Oct 2006

    Myth, again

    OK, I can take a hint. The internal Intel display chip is too fussy, and I have wasted way too much time on it. Also, it appears that you can't just take a disk out of a Dell and drop it into a Compaq. It says something about loading a table, and then demonstrates the halting problem.

    So the Dell, with its useful quiet fan and 2.4GHz CPU and useless display adaptor, is back mostly where it started, at least with the wires untangled. The noisy AMD 1.7 GHz (if I named all these computers, I'd start sounding like Jerry Pournell. Sure wish he'd come back to Hackers.) is sitting opened somewhere, and a Compaq 1.7MHz Pentium 4 has the frontend system. Undid all the tests, and it seems to be working fine, though I haven't checked the 1080i yet. That requires that I recompile myth with XvMC turned back on and enabled. I think lack of that was giving me the jerkiness in the first place. The Compaq is MUCH quieter, which is good for a front end. It also came with a less-moosy NVIDIA card which seems to be just fine with the Plasma screen. I've had trouble with Compaqs and Unix before: too much special firmware and such. But this one seems to be OK.

    So I am recording some NASCAR (that HAS to be hi-def, right?) and recompiling myth, and in 50 minutes or so, the test should be ready. And the piles of open computers are shrinking a bit.

    We had a 14-hour Internet blackout last night. It was Verizon's fault, again. Near the top of my list is checking out the home network connection options. We are paying $165 for 768/768 and a /28 network. That's too damn much for not nearly enough. The cable guys will get their chance, and with their latest, and Verizon, which is busy bringing FIOS to Bernardsville. It's already at the bottle of the hill, though I don't see it on our polls. Talked the the Verizon guy: he said the whole town will be wired and running video by the end of the year. All parties have better speed and pricing. It's the static IP and net block that is unusual, and so maybe not on their price list. Of course, I could do some interesting asymmetric routing of packets, something else they probably haven't thought much about.

    In any case, I have to get the dialup backup working, and get that link to Brian working. I have the time to, now.

    Anyone know how to use the psm0 port as a standard serial port, connecting to a modem? Off to google...

    13 Oct 2006

    Myth, again

    ...and it is never easy. The problem is jerkiness at the highest video rates, which is to say NBC (1080i) on Earl night. I've seen this a couple of times before, but it has always lasted a couple seconds, and went away. Clearly, an important factor is that I am running a 1.7GHz Athalon with NVIDIA card. 1.7GHz is on the hairy edge of underpowered for this. Deinterlace selections (especially OFF) and the mpeg library used matter. I even cranked the CPU up: 105% (ah, I remember running margins on the old CDC equipment) works, but not enough cycles, and the machine crashes during boot at 110%.

    We have a 2.4GHz Celeron (a Dell) in use elsewhere. VIA chips aren't a factor in this front end-only machine, it is much quieter, and the black is nicer in the family room. Let's swap them. Seems easy enough.

    Well, it's never easy with Myth. The Dell doesn't have an AGP slot, so I can't just drop in the NVidia card. Everything would Just Work (I think) it I did that. In retrospect, the most cost effective solution would be to purchase a PCI NVIDIA card, but I don't know if they are available any more. We do have a number of oldish display adaptor cards, including some for PCI. But here's the rub: our plasma display is a Panasonic TH-50PHD6. We love it. It's top resolution is 1366x768. As we learned with our 20 inch Dell LCDs, a lot of graphics adaptors, especially older ones, will not switch to non-standard scan rates. No problem, you say: I am running Linux (still FC4) and Xorg, and we just give the desired modeline in the xorg.conf page. And we use the Intel 845 chip on the Dell motherboard, which isn't bad. Even has 64MB of video RAM. Should be good enough.

    Two problems with this: 1) the unmodified Intel will not switch to resolutions not found in its BIOS. 1024x768 is there, and several other of the usual suspects. But the drive ignores modelines that aren't in the BIOS. There's a program on the net, 915resolution, that changes the table in the Intel RAM copy of the resolution table. So I put in 1366x768, and now it complains about problem 2:

    *(WW) (1366x768,plasma) mode clock 88MHz exceeds DDC maximum 0MHz
    I am actually wondering if the chip won't crank out 88MHz, though that doesn't seem like it is too bad.

    The display works fine at 1024x768, but myth is complaining about something new:

    Xlib:  extension "GLX" missing on display ":0.0".
    Xlib:  extension "GLX" missing on display ":0.0".
    Xlib:  extension "GLX" missing on display ":0.0".
    QGLContext::makeCurrent(): Cannot make invalid context current.
    and on the console:
    Oct 13 18:23:51 atv kernel: audit(:0): major=252 name_count=0: freeing multiple contexts (1)
    Oct 13 18:23:51 atv kernel: audit(:0): major=113 name_count=0: freeing multiple contexts (2)
    Cheswick's Rule: If you have a problem you don't understand, and a message somewhere else that you don't understand, go fix that other problem. It may be related.

    I'd look up these messages, but the DSL line is down, so I am writing this instead. And I still don't know if the Intel onboard display chip has got enough moxie, along with the 2.4GHz CPU, to actually fix the problem.

    Meanwhile, the papers for NDSS are done. I did one review as a sonnet. (They started it.) Doggerel---I've never been any kind of poet---but a fun challenge, and I liked the paper.

    Tommorrow actually is Saturday. I've had about 13 in a row so far, and piles are starting to diminish, things are getting fixed, and the vast reading piles are shrinking. And I am getting 45-75 minutes of hardy exercise almost every day. So far so good.

    10 Oct 2006

    Four NDSS papers to go. Done tomorrow, for sure!

    Cool Science

    We know they've searched for new antibiotics in soils from around the world for decades. These are simply tools in the arms race in dirt. I just read that now they are looking for antibiotic resistance in soils, and have found lots of new methods of resistance. Normally, the bugs get resistance to our antibiotics by single nucleotide mutations. These are whole new methods and enzymes not known to our personal denizens.

    D. radiodurans is highly resistant to heavy doses of radiation. Where did it learn this trick? These levels of radiation are not found in the environment. It turns out the dessication causes the same kind of DNA damage as radiation. Cool. (Nature 443, 569-573 (2006).)

    5 Oct 2006

    Today was spent reviewing an NAS committee report. These things get a bit turgid at times, but it is a good report. I hope my comments are useful.

    Entertainment has varied a bit over the past few days. We saw Richard III at Kean on Sunday, a play I had never seen. It was nicely done. I hope they are doing Lear sometime soon.

    One of the things I am doing during this break is taking time to fix things in my computing environment that annoy me. I am inordinately pleased to learn that adding

            ForwardX11Trusted yes
    has fixed a long-time problem of lot X connectivity, and failure of the sam "exch" menu option. Everytime it doesn't fail, I am delighted. I should ahve taken the five minutes to hunt down the actual problem years ago.

    I miss the General Research Colloquium at Bell Labs, which died of the insolvancy of Lucent. I am now trying to feed my hunger for good lectures on the web. Princeton has a variety I want to see, so I started with part one of a three parter on String Theory by David J. Gross. Part 1 last night cleared up some confusion I didn't know I had about QCD. I am looking forward to the remaining two parts, and the economics and judicial lectures as well.

    Of course, there are fascinating lectures available from institutions all over the world, and it would be nice to collect them (perhaps through Google?) and make them easy to see. (Perhaps this has been done already: I will have to check.) I think MythLecture would be a great idea, and I might even do it myself over the coming weeks. A problem I see is that these tend to be available via Realaudio or Mediaplayer, and I am not sure that either format is available on Linux. Hmm. I would be willing to collect and moderate such a list of lectures, at least the ones I am interested in, and can imagine that it might be a nice wiki.

    The new "instahot" hot water heater is great. I am drinking more green tea.

    Got to get today's exercise, then dinner, then part 2, then Earl.

    4 Oct 2006

    Saturday Night Live has started a new season, and apparently they have new writers. Several of the sketches were actually funny, quite a change for them. I guess you can teach an old Lorne new tricks.

    The endless Saturday continues, and I am very busy. Not only are there a huge backlog of Saturday chores, but I have 20 NDSS papers to grade, an NAS report to finish reading and commenting on, a written interview to finish, and I spent 45 minutes on the phone with Forrester discussing Internet security.

    There seem to be a lot of decent papers for NDSS this year. Last time I was on the PC for NDSS, I read over 20 crappy papers and no good ones. Only chasing down the references gave me any respite from lousy papers and trivial ideas. Better than 50 percent seem to be at least plausibly good papers.

    I meant to review five a day, but three seems to be the right match of reading, research, and other chores. It certainly is a good way to stay up-to-date.

    Through all this I am continuing to exercise every day for at least 45 minutes. Walked to the grocery store this morning, 50-minute round trip and appropriate that I am hauling food. Along with schedule planning and reading, that's most of my day these days, until the papers are done and I can get down to the science museum exhibit software and cleaning out this cluttered up home.

    Latest JAMA had several articles on obesity. It's clear that if you are insulin-resistant, a high-fat hat-protein diet is better than carbs. Bought burgers today. At this point, travelling to British soil to score some rimonabant is plan B. A couple months of this exercise regimen should do wonders.

    29 Sep 2006

    I cut out of work early today because why not. Following a farewell Arby's and a nap, I proceeded to the first project:

    Autopsy of an In-sink-erator Model GN-3B-10 1/2 gallon hot water heater
    It is already established that this is a POS. Why? Because it died within five years of installation, which offends the engineer in me. Also because I was unable to find any manuals whatsoevery on their web pages about the device. And because we never got a call back from 1-800-558-5700, the service number we called several times.

    The thing started to leak, apparently from somewhere inside the sealed box (no user servicable parts: another strike against it). So I took it off, popped the rivets, and checked it out.

    The autopsy is inconclusive. There is no obvious leak source inside, though there are places I did not go---there are too many sharp edges. If one of the connections on the top leaked, there is no sign of it in all the mineral deposits I've found.

    The question is, what do I replace this thing with? One of the goals of this retirement is to catch up on numerous projects, many of which are of the research-it-and-buy-it variety, which can take a lot of time. Googling for reviews of these instant-hot-water devices has been disappointing. Do I buy a fine German brand (something I would never do for a car)? Go to a store and inspect their wares? That's probably what I have to do. At least I'll get the tea flowing soon for retirement.

    Previous entries