Older blog entries for thomasd (starting at number 12)

Don't seem to have got very much done this week -- possibly due to having to give several presentations in quick succession. I guess the good side of this is that I should have to present anything else for some time to come. Or at least, that's the theory.

Software: Just got a new build of Mozilla running on my alpha box, and it rocks. Last month, it was dying horribly in hard-to-pin-down 64-bit-only ways. Now it seems as stable as it is on my Linux/intel machine. Well, I guess this means the end of my happy (yeah, right) relationship with Netscape Classic.

I've also been experimenting with a development snapshot of PostgreSQL. Specifically, the new 7.1 feature which finally removes the block-size limit for table rows. I can now happily dump hundreds of kilobytes of genome sequence into normal `text' attributes with no fiddling at all.

Coding: Found a bit more time to work on the Java EnsEMBL API. It's now connecting to the database and I can fetch a few object types. Hopefully have it complete enough to be useful for me in the next few days, then I can concentrate on getting it ready to demonstrate to the other EnsEMBL developers. Guess it'll be a while before the project really moves away from Perl, though.

Excursions: Went to the Imperial War Museum on Saturday. Lots and lots of aeroplanes (plus various other kinds of vehicle). One of the exibits was an old (late 1940s, I think) radar guidence system for an anti-aircraft gun. How did they do that without microprocessors? If it were designed today, you can bet that it would be running NT or somesuch.

WebDAV: Had a browse through the specs (plus the delta-V versioning protocol), and it looks good. Looking forward to subversion even more now.

After a lot of digging, and a recompile of gcc, I think I've finally got to the bottom of the Java/C++ on Tru64 Unix problem. If you're using threads, you have to compile absolutely everything with _REENTRANT defined. I guess that makes sense, but when there isn't a reentrant version of the standard C++ library, it doesn't make life easy... Wish I could stick to Linux.

Homework: Read up on WebDAV.

Spent much of the day trying to get Java native methods working reliably on Tru64 Unix. Conclusion: if you want to use C++ and iostreams, forget it.

Have been looking a bit at subversion. I've been using CVS very heavily for the last year, but this looks really promising. Might try grabbing a copy over the weekend to see how development is going.

Added exception support to the bytecode library. Still think the .class file format is wierd, but I've got quite a few of the wrinkles hidden behind nice library code now. Ought to get round to writing some web pages explaining what's going on.

Gradually making progress with the EnsEMBL API.

Largely been focusing on my research projects for the last few days. Some interesting new strands coming together -- hopefully find the time to write it all up some day soon.

Friday 13th today -- wonder how my luck is going...

Coding:Talked about the design of a Java API for EnsEMBL. I'll start coding this up over the weekend, hopefully, and it might reduce the chances of me having to write more Perl.

BioJava: Landed autogenerated feature proxies today. These are the first thing to use the new code generation module, and I'm appreciating it already.

It's probably time to start a proper TODO list for BioJava 1.1. Ouch.

Databases: Read about the plans for long data support in PostgreSQL 7.1. Wheeeeee!

Think I've got the .class file format sussed now. Or at least, I can generated classes which validate and run correctly, so if I haven't got it right, we really should be worrying. The library is still missing a few bits and bobs, but it's complete enough to unblock our work on BioJava.

Real life: Beautiful, sunny day -- you really wouldn't tell it was October. Now why didn't they listen to my plan for wireless Ethernet in the gardens? *sigh*.

Re: scripting languages. Nice to see some support for Tcl, even if my gentlemanly credentials are at stake here. I quite agree about treating it as a C library. It's also good for scripting Java objects using TclBlend or Jacl. For tasks where there's a good Java class library, this combination lets you get a script up and running very quickly.

Had full-on exposure to the Sun .class file format today. I've always been reasonably happy with most of the bytecode itself, but haven't had much reason to look at the whole file format until now. It seems to be quite heavily tied to the guts of Sun's classic VM. I guess I now have a better idea why writing a cleanroom VM is such a major undertaking. The schedule for the bytecode library has slipped slightly since my parents turned up in Cambridge, but maybe I'll get it finished tonight.

Real life: looking around for something a bit more active to do (but preferably not too much running round in circle). I think I might give fencing a try.