Work Blog

Moved

ttroxell — Mon, 15 Mar 2010 04:31:06 GMT

I don't think I will use this much anymore. Go here for new posts.

Or-ing procmail rules

ttroxell — Tue, 30 Jan 2007 19:56:58 GMT

Is there no better way to OR procmail rules than these horrid kludges?

For now I'm going with the DeMorgan Rule method, because it's more efficient and to the point that the alternative kludges. I still feel a little dirty writing things this:

      :0
      * ! condition1
      * ! condition2
      { }             # official procmail no-op. MUST LEAVE SPACE
      :0 E
      action_on_condition1_or_condition2

It would be nice to have a recipe flag that means "OR the conditions below"

De-Scheming MIT

ttroxell — Sun, 19 Nov 2006 12:03:57 GMT

I am amazed by MIT's decision to move it's EECS curriculum away from scheme, doubly so at the fact that they're swapping it for Python. It has been part of their definition, at least in my mind. It makes sense to me, but I suppose I have some bias. I see it as a move from a language with deeper cognitive benefits to a language with stronger pragmatic benefits-- a move towards actually building things. Of course, it's just a language, and the same course could be taught with either one.

/~xtat/debian-security-talk06

ttroxell — Wed, 04 Oct 2006 23:31:59 GMT

Debian Security talk went well!

I broke my glasses about 2 minutes before starting and a kind Lehigh student (engineer?) repaired them--- with a staple? Thanks, whoever you are!

My 150 (!) slides are here: PDF, ODP, SWF.

Free as in Freedom

ttroxell — Mon, 04 Sep 2006 09:00:25 GMT

I spent some time this weekend reading Free as in Freedom: Richard Stallman's Crusade for Free Software, and found it quite enjoyable. It is anecdotal, much of it reminding me of folklore.org. Regardless of what you think of RMS, it's worth a read, and might just change your mind. (Full text is available here.)

logcheck repository move

ttroxell — Sun, 09 Jul 2006 10:31:03 GMT

Thanks to Martin, the logcheck repository has finally moved from CVS to SVN. Hurrah! not CVS!

USPTO/ US Homeland Sec and OSS

ttroxell — Wed, 11 Jan 2006 22:21:04 GMT

"Through its Science and Technology Directorate, the [US] Homeland Security Department has given $1.24 million in funding to Stanford University, Coverity and Symantec to hunt for security bugs in open-source software and to improve Coverity's commercial tool for source code analysis."

Link

"The United States Patent and Trademark Office plans to announce today that it will cooperate with open-source software developers on three initiatives that it says will improve the quality of software patents."

Link

Two optimistic things in US Gov't.... (It's not April 1 yet, right?)

>_<

ttroxell — Mon, 28 Nov 2005 19:48:04 GMT

We're Sorry...

You seem to be viewing the system in more than one window. This functionality is not currently
supported. Please close this browser window and continue working in the second browser window.

(In the future, if you need to have more than one browser window pointing to the system,
please make sure you DO NOT launch a new window from an existing window, using -N or from
File -> New -> Window. This can cause the windows to interact with each other. Instead, please
launch a separate window by double- clicking your browser icon. Thank you.)

...and thank you!

for having awesome restrictions like this on your web app.
for not telling me this in advance, forcing me to retype 2 paragraphs.
for not giving me any way to navigate away from this page, and breaking the back button.

dumb asses

ttroxell — Sat, 19 Nov 2005 10:27:19 GMT

Found this under heading "What is a statement of verifiable fact" while browsing EFF's Legal Guide for Bloggers

A statement of verifiable fact is a statement that conveys a provably false factual assertion, such as someone has committed murder or has cheated on his spouse. To illustrate this point, consider the following excerpt from a court (Vogel v. Felice) considering the alleged defamatory statement that plaintiffs were the top-ranking 'Dumb Asses' on defendant's list of "Top Ten Dumb Asses":

A statement that the plaintiff is a "Dumb Ass," even first among "Dumb Asses," communicates no factual proposition susceptible of proof or refutation. It is true that "dumb" by itself can convey the relatively concrete meaning "lacking in intelligence." Even so, depending on context, it may convey a lack less of objectively assayable mental function than of such imponderable and debatable virtues as judgment or wisdom. Here defendant did not use "dumb" in isolation, but as part of the idiomatic phrase, "dumb ass." When applied to a whole human being, the term "ass" is a general expression of contempt essentially devoid of factual content. Adding the word "dumb" merely converts "contemptible person" to "contemptible fool." Plaintiffs were justifiably insulted by this epithet, but they failed entirely to show how it could be found to convey a provable factual proposition. ... If the meaning conveyed cannot by its nature be proved false, it cannot support a libel claim.

Yahoo 2.0

ttroxell — Thu, 10 Nov 2005 12:41:17 GMT

looks like Zawodny has made the same conclusion about Google that I wrote about earlier. Re: moving in on existing markets rather than creating new ones. Like I said before, isn't that what they've always done?

Calling Boston Debian Gurus!

ttroxell — Wed, 26 Oct 2005 20:39:16 GMT

If you're interested in doing sysadmin work in a heavy Debian/Ubuntu environment in the Boston area, check out this posting for a Senior Network / Systems Administrator.

Logcheck Wiki

ttroxell — Sat, 15 Oct 2005 16:18:37 GMT

Eric Evans has kindly set up a wiki for the Logcheck project. Some braindump-style notes on the upcoming macro system are on the LogcheckTemplateSystem page.

babygrep

ttroxell — Tue, 11 Oct 2005 01:24:09 GMT

I made a baby grep.


sushi% ls -lh babygrep
-rwxr-xr-x  1 xtat xtat 3.6K 2005-10-11 04:10 babygrep
sushi% ls -lh /bin/grep
-rwxr-xr-x  1 root root 70K 2005-03-10 18:09 /bin/grep

I am wondering if using something similar would make logcheck faster. It needs work in any case. I just think it's cute.

code

Logcheck Rulefiles Analyzer

ttroxell — Mon, 26 Sep 2005 23:31:31 GMT

I spent some time making a rulefiles analyzer for Logcheck.

Noticing that logcheck-database had 1000+ rules made me curious about utilization. This tool will report the number of matches for each individual rule. The output looks like this right now:

*cut*
file: rulefiles/linux/ignore.d.server/dhclient:
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 124, 124, 0, 0, 0, 0, 0, 0, 124]
file: rulefiles/linux/violations.d/logcheck:
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
file: rulefiles/linux/ignore.d.server/policyd:
[0, 0]
file: rulefiles/linux/ignore.d.workstation/winbind:
[0]
file: rulefiles/linux/violations.ignore.d/logcheck-cyrus:
[0, 0, 0]
file: rulefiles/linux/ignore.d.paranoid/cron:
[0, 0, 18, 0, 0, 0, 0, 0]
file: rulefiles/linux/ignore.d.server/nscd:
[0]
*cut*

The numbers in the arrays correspond to line numbers in logcheck rulefiles. This output will be improved eventually. It should also calculate the top N and bottom N matched rules. This should aid us in tracking down stale rules.

analyzeRules - Get statistics about logcheck rule effectiveness

You'll need python 2.4.

More on acoustic keyboard sniffing

ttroxell — Tue, 20 Sep 2005 15:16:12 GMT

10 minute recording, 96% recovery of typed characters. Keyboard Acoustic Emanations Revisited. via CleverCS

ttroxell — Fri, 12 Aug 2005 01:56:36 GMT

John Markoff's new book sounds interesting: What the Dormouse Said: How the 60s Counterculture Shaped the Personal Computer
link

ttroxell — Sun, 10 Jul 2005 06:41:58 GMT

I interfaced our Daily Nav Report with Google Maps using XMLHttpRequest and some client-side parsing.

The result is a Google Maps based ship tracker.

It's just a demo for now.

openssh continues to impress

ttroxell — Sat, 25 Jun 2005 14:01:10 GMT

OpenSSH can act as a SOCKS4 proxy directly! I had no idea. It is quite handy for encrypted AIM if you don't trust your intermediate link, for example over open wifi, or in my case, an untrusted satellite network.

Anyway to get it set up add "DynamicForward 1080" to ~/.ssh/config
That's it. Config your gaim to use localhost:1080 as it's SOCKS4, ssh to remote host, connect with gaim.

You probably want to do it from the command line each time if you're often connecting to untrusted hosts.

(but I still don't like apple)

ttroxell — Thu, 16 Jun 2005 11:46:13 GMT

I caved in. I have an iPod. It contains a full i386 Ubuntu mirror. This will be useful while at sea.

Hfsplus on hoary is buggy. It seemed to work at first, but then when I later remounted the iPod, kernel said that it was previously unmounted uncleanly and that I should run fsck.hfsplus. I tried the hfsck utility from the hfsplus package and it exited fatally without fixing anything.

If you're going to try this, be sure to format your iPod with FAT32 using the cd that came with it if you expect any sort of data integrity. The windows version will do this without asking.

The Art of Computer Programming Reading Group

ttroxell — Wed, 18 May 2005 04:09:06 GMT

Josh Myer has started a TAoCP reading group to cover all three of Knuth's volumes in a year. Mailing list here.

OpenID

ttroxell — Wed, 18 May 2005 02:22:53 GMT

Brad Fitzpatrick of Livejournal fame has started an interesting new project called OpenID which calls itself a "distributed identity system."

Think Passport, minus the evil and unpractical bits.

Or for a better summary:
"Logging in to a dozen websites every day is lame."

eek

ttroxell — Thu, 12 May 2005 22:36:35 GMT

Hyper-Threading considered harmful?

ERR_NOENT

ttroxell — Tue, 03 May 2005 11:13:18 GMT

Things I learned in 16.5 hours at the lab yesterday:

Linux NFS works for most everthing but Solaris; probably because Sun insists on following standards instead of making things interoperate in any practical sense. See Redhat #144556. We had to revert our change to a Linux NFS back to Solaris because of a few folks who still have old Sun boxes. The whole situation felt very Microsoft. It is only a temporary solution.
Solaris boxes DEADLOCK on trying to mount an NFS volume at boot if the entry is incorrect or the server is unavailable. I've seen similar behavior on other Unix, but usually it times out eventually.
Redhat 9's[1] minicom will SEGFAULT and break your serial port until reboot when you perform certain operations in a vi session to a Solaris box. This is especially cool when you are running it from your web server.
I wish Solaris would disappear immutably.

Unrelated: Why Do Computers Stop and What Can Be Done About It? Jim Gray, 1985

[1] Yeah, I know. I need to replace that old beast.

desk-lighting user interface

ttroxell — Mon, 02 May 2005 15:30:42 GMT

I just thought of an interesting lighting UI for desks.

I would like desk lights that have infrared sensors to detect my presence at the desk-- and not turn on suddenly, but increase the power to become brighter gradually over a second or two. As soon as I stand up, they should [just as slowly] return to the very dim state.

It would be really cool to have this behavior in a lab of computers.

I am most comfortable working in near-darkness or twilight, and I think it would be even better in this situation.

It doesn't seem like it would be that expensive to set up... I bet the green folks would be interested in the power savings.

fdata-intersect

ttroxell — Sun, 01 May 2005 19:54:25 GMT

Just 'cause I felt like using Google-goopy:

mata% ./fdata-intersect.py xtat myagi
processing xtat...
processing myagi...
intersection of 2 friends for xtat and myagi:
errric
xtat

fdata-intersect.py