From: Neal R. <ne...@ri...> - 2004-05-01 22:19:39
|
> This may sound like heresy, but should we back-port all of the useful > features of ht://Dig 3.2 into 3.1 and give up on the 3.2 branch? Joe > showed a factor of 4 speed difference, and today I was getting a > factor of over 10 difference (after much tweaking of the 3.2 code). Interesting... I guess it's time to do some heavy profiling. On the plus side of 3.2.0 is that we're tested it at RightNow very well.. very few bugs so far. > I'm not sure who wrote the new database code, but they seem to have > left the team and no-one seems to be jumping up to pick up the > pieces. I certainly don't have the knowledge, or the time to learn. > For a while, Neal seemed to have the interest and skill needed for > rewriting the code, but he seems to have found greener pastures too. > (Are you still with us, Neal?) Yes I am.... I'm working on HtDig code (or code that uses htdig) frequently... sorry I've been quiet... working on a PhD part-time and working full-time... > Does anyone know what new features require the new database structure? > Phrase searching is the big one, and compression is the other one > which will take a lot of work. There's also the ability to set > weights at query time rather than dig time, but that is presumably > just a matter of replacing a "weight" field with a "flags" field and > keeping the rest of the database structure unchanged. > > In addition, 3.1.6 builds much faster, and doesn't use automake (which > I would like -- those tools seem a nightmare). > > My vote is that we release 3.2.0b6 basically as the code stands now, > and then start 3.3.0a1 by back-porting features to 3.1.6. For each, > we'll measure the impact on performance, and decide which ones are > worth it. I'd agree with the first part, but not the second until we try and figure out why the code is slower. Neal Richter Knowledgebase Developer RightNow Technologies, Inc. Customer Service for Every Web Site Office: 406-522-1485 |