« Jeff's Journal | Main | There is nothing more joyful than wallpapering behind a toilet »

I'm the original Nicks Superfan

My IP traffic to IU goes through Atlanta, then to New York, then to Bloomies.

My pilot fritzed out again the other day and reset itself for no reason. Last week, it fritzed out and lost all of its data (also apparently for no reason). This wasn't a tragedy, because I had done a backup in the recent past, but I did end up losing some data, which was annoying. This pilot is just getting kinda old; it was pretty hard to write on it (letter accuracy was pretty bad). Not that I'm an expert at palm writing, but when I wrote on Tracy's (new) pilot, my accuracy is much better. This leads me to believe that my (old) pilot is just getting tired. Indeed, the writing area is visibly worn.

So I went out an out a new m100. There's lots of other more advanced models -- indeed, m100 is the low-end offering -- but that's really all that I need. Although I had visit three stores before I finally found the m100 in stock (at CompUSA). Best Buy and Staples were both sold out.

I transferred over all my data and everything appears to be find with the new m100. My old pilot has officially been retired. It served me well.

So this brings the tally to five -- this is my fifth palm pilot. The record stands:

Pilot number Status
1 Death by rollerblading
2 Lost when left in the back seat of a rental car in Florida
3 Death by 3 foot drop onto asphalt
4 Retired at age 72 (palm pilot years)
5 Still in service

Tracy has broken a pilot as well (I think it was a short drop from her desk...?); she's on her second pilot. Now we both have identical m100s. Ugh...

I bought Depeche Mode's new "Exciter" CD while I was in Bloomies on Monday (stopped in Target to buy some Bennadryl, and ended up wandering over to the CD section). It's not bad, but the music is slower than their previous stuff.

Had a really long SSI call today. Dog, Brian, and I hammered out a whole bunch of stuff -- we've had separate discussions up until this point; this call helped have all three of us agree on a whole bunch of points. There's still a ways to go, but we agreed on things like:

  • There will be different kinds of modules (e.g., a comm module for ipv4 vs. ipv6, a boot module for the lamboot kinds of things, etc.).
  • Some kinds of modules will be use-only-one-of-all-available-modules (e.g,. comm), whereas some kinds of modules will be use-all-available-modules (e.g., the RPI).
  • The SSI glue will be general enough to not care where the modules are located (in liblam or libmpi).
  • The SSI glue will have three essential functions: init, finalize, and export_tables.
  • LAM's top-level configure will write out the tables that the init functions will use
  • Each module will have its own top-level interface that the rest of LAM/MPI will call. These top-level interfaces will do their own dispatching.
  • Modules will emulate C++ inheritance by specifying two global variables in each module: a struct full of a bunch of function pointers, and a pointer to a "base" struct full of a bunch of function pointers (which can be NULL).
  • Modules with NULL values in their struct of function pointers will have those NULLs replaced with real function pointers from their "base" class during init, so that during run time, we only have to do one pointer lookup, not [potentially] many.
  • The SSI finalize glue may not be called, but only if the process doesn't call kexit() (e.g., the lamd dies by signal)
  • The init function of each module may fail if that module determines that it should not run.
  • There will be command line interface that may be used to force the selection of a given module. It will likely have a three part nomenclature: a common option, the module kind name, and the specific module name.
  • Modules can call the SSI export_tables routine to get the final function pointer table for a given kind (probably with arg type (void*)) so that the SSI dispatch wrappers can be bypassed for performance reasons (this is important for the RPI).

I think those were the majority of things decided, in addition to the function breakdown of several of the [proposed] kinds of modules. One notable thing that we haven't decided yet is how to handle arguments of different types. The comm module has a notable problem: argument types and sizes are different between ipv4 and ipv6. We may get away with using handles to the "real" datatypes in many places, but Brian thinks we're going to have many problems in the routerd in the lamd, because it stores tables of IP addresses. Ugh. We'll see how that shakes out...

Connectivity to ND has sucked today. It keeps going away for 1-2 minutes at a time. Very frustrating...

I got /. moderator access this morning; I don't think that I've ever had that before. So I moderated a 5 articles and did my civic duty.

I went to see Tomb Raider with Tracy last night. It was fairly good. It wasn't what I would call a great movie, but it was worth seeing on the big screen and whatnot. Good effects, but I found some of the action a little hard to follow because they kept switching the camera at a high frequency. I'm sure that they did this on purpose, but I didn't like it much.

All in all, though, it was a good flick. Rimmer (Chris Barrie) from Red Dwarf was the butler. I give the movie 15 minutes.

Got some LAM patches from the KLAT folks to make LAM work on their systems. Their networks have the property that each node has multiple NICs, and the switches are wired up in a non-uniform way. i.e., nodes A and B may have entirely different ideas of what the IP address of node C is.

So LAM has to do something a little different than what it normally does: pass around hostnames rather than host addresses. Tim from the KLAT project hacked this up in LAM and sent me a patch. I'll probably be applying it soon.

Tim also noticed a small bug in LAM, such that when /tmp is NFS shared across multiple nodes, when the haltd in the lamd dies, it leaves the kill file open. This causes NFS to keep a cache file open (of the form .nfsNNNNNNNN) in the LAM directory, and therefore doesn't let tkill remove the entire directory. Blech. Tim's fix for this doesn't work when one compiles the lamd as separate pseudodaemons, so I passed it on to Brian to see what he can do with it.

I renewed squyres.com today for another 2 years (thanks for reminding me, palm pilot!) I renewed it a few months early, but hey... I don't wanna lose it. :-)


I have it on reliable sources that Nabisco is considering releasing a new animal for animal crackers since this year is the 100th anniversary of animal crackers. There are four animals under consideration; right now, Nabisco employees are being asked to vote for which they like the best. Supposedly, Nabisco will open the voting to the public later this year.

Here's the four animals under consideration:

  • Koala bear
  • Walrus
  • Cobra
  • Penguin

I think we all know what it should be. When Nabisco goes public with this, we'll have to post this to Slashdot and get all those Linux lovers to stuff the ballot box.

ND was supposed to switch to kerberos 5 this morning. By our syslogs (and the 6 bazillion cron e-mails that I got), AFS was out (on schedule) for several hours this morning, and came back a little after 6am. But checking several machines after they have been rebooted (they needed to be rebooted to get the new authentication scheme --
it's complicated), they didn't appear to be using kerberos 5.

What's the scoop?

Not surprisingly, the OIT has said nothing. Curt sent a "WTF?" kind of message to the AFS/Unix list, but AFAIK, there's been no response yet (it's about 2pm now). The OIT really sucks sometimes. Haven't they ever heard of communication? Haven't we been harping on their [lack of] communication skills for years now? Haven't they learned yet?

Apparently not. :-(

I just got paid by the military for my 2 weeks in Atlanta, but DFAS took out state taxes for IN even though I explicitly submitted a form saying KY. <sigh>

Some guy just posted a Solaris system administration question to the LAM list. Weird.

Post a comment

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)


This page contains a single entry from the blog posted on July 14, 2001 2:07 AM.

The previous post in this blog was Jeff's Journal.

The next post in this blog is There is nothing more joyful than wallpapering behind a toilet.

Many more can be found on the main index page or by looking through the archives.

Powered by
Movable Type 3.34