|
April 9th, 2008
01:39 am - office move
We've been staying at the same office building for almost 20 years (we moved suites once during that time), but finally we moved a bit closer to the LAX airport. I hate moving.
We lost power to the half of the machine racks today. It turns out that the poor budgeting of load per circuitry caused one of the circuits to the machine room to overload. As I was not involved at all in the actual move operation, I cannot complain. Other people did the hard part of the job and I should just be thanking them.
|
August 24th, 2007
02:52 am - Bugfix
Had to spend embarrassingly long time in the day job to hunt down an obscure bug, which turned out to require an exactly one byte addition to the code to fix. Wrote about 20 lines of log message describing the change.
|
July 24th, 2007
11:40 pm - Wrapping up Socialtext contribution
I've been working with Socialtext folks for about three weeks, helping them to add tokenization of Japanese text for indexing and searching in their Wiki software. People from NEC greatly helped in testing every little details, and I think I am done.
I have been writing my day-job code in a language I did not quite like (well, that might be an understatement), and Socialtext work, whose primary language is Perl, as a side project was quite a nice change.
|
July 10th, 2007
04:45 pm - Still a bit more SocialText
Today, I think I went a bit overboard. There are quite a few idiosyncratic characters in Japanese that are not portable and simply ignored by popular language analysis programs. E.g. you could write your list like this:
① do this thing
② do that thing
③ realize the last emperor's era was 昭和 but there is an idiosyncratic variant ㍼ (mind you, this is two characters squashed into one and given a single codepoint. What a mess).
For the purpose of building search index, these "circled numbers" can simply be ignored (looking for ① would be a moral equivalent of looking for pages that contain (1) in English text). But it would be nicer to unsquash some of the "Squashed" letters, like above.
Unfortunately working on this was quite painful, as the version of Emacs I use everyday does not grok these idiosyncrasies (that is why they are called "platform dependent characters").
Current Mood:  tired
|
July 9th, 2007
10:51 pm - A bit more SocialText
I earlier gave some code to SocialText folks, who in turn had an NEC guy review it. Extra set of eyes always help, even when a reviewer is from a different background of programming. Among what he came back with were a handful of good comments and suggestions.
I ended up rewriting a small portion of the code, which I did not mind at all. Integrating the result into their build infrastructure turned out to be a bit cumbersome for me (not that their infrastructure is bad in any way; just I am not familiar with it), but they offered to just take the patch from me and do the integration themselves.
Which was very nice of them. It obviously is the right thing for both of us (after all, it's their build infrastructure and it would be trivial for them, than for me to learn the subtleties of it while keeping them waiting) from technical point of view, but I haven't seen that kind of "you'll help us there, we'll help you in the areas where we are more familiar with" in business settings. I always seem to hear that "we asked you to do all of that, so please follow through all the way to the integration step, we don't do your work". Not with SocialText folks. Very nice.
Current Mood:  pleased
Current Music: Chico Buarque
|
July 3rd, 2007
12:46 am - Silly customer requirements
Sometimes, silly customer requirements force one to write a stupid hack that one does not believe in, and today was one of those days. I ended up coding a snakeoil copy protection for a commercial product.
A funny thing is that I can get away with a snakeoil security in this particular case because DMCA protects it. That does not make me feel better, of course. But a day job is something one does, and does it well if one must. One does not have to like it.
Current Mood:  embarrassed
|
June 29th, 2007
10:26 pm - Day job again.
I usually have a bit of quality git time during lunch hours, but somehow I've been robbed of that this week, which is not a very good sign.
But fortunately the other guys in the day-job project are picking up. Maybe I can take a break soon.
Current Mood: tired
|
June 28th, 2007
09:51 pm - Back to day job
Since SocialText folks seem to be either attending Perl conference or taking a vacation, I haven't heard back from them on the changes I gave them yesterday. Not that I expect any major issues with them, though...
So today I switched back to the other day job project, that I am too embarrassed to talk about much. Japan (where we have our sales side of the project planning people in) asks fairly large last-minute additional development, but I am not surprised anymore X-<.
Current Mood:  grumpy
|
June 27th, 2007
11:39 am - Having fun with SocialText
My employer is helping NEC (from Japan), and they are involved in localizing an OSS Wiki project SocialText to grok Japanese. The software is heavily on Perl (which is fine by me), the people seem to be very high quality (again, fine), and not hesitant to be on the bleeding edge.
Somehow I got sucked into that project, and ended up writing an interface module to run morphological analysis on Japanese text for indexing purposes.
Their main repository is in SVN, some parts of which are protected with authentication and others are public. Unfortunately git-svn seems to be having quite a lot of trouble when dealing with their branches (the trunk import went fine). Tentatively I've given up using git-svn to maintain my part of changes, and using raw git over SVN checkout instead.
Today was the first day I actually mucked with their development environment. My code has been tested on my private machine so far, so I had a fair amount of confidence in it, but it certainly is assuring to see that it worked as expected on somebody else's environment.
|
|
|
|
