Hi, At 14:40 02-02-07 -0600, Gilles Detillieux wrote: >According to Robert Marchand: > > Yes all you are saying is true! I was reluctant to test the snapshots > > because I tought It would be trivial to pass from 3.1.5 to 3.1.6 and the > > only reason I wanted to install 3.1.6 is for future security fixes who > will > > be I suppose for that version. Everybody is happy with Ht://Dig here > so we > > don't need (as far as I know) any new features. > > > > You know, even if I had tested with the snapshots, I'm not sure I would > > have see the problem. I found it because I replace the CGI on the live > > server and It made a mess pretty quickly (the load was over 40). The live > > server has around 80000 documents and it's an old machine. On the test > > server (with 2000 documents), I didn't saw the difference at > first. The > > test server is a workstation so I tought it was OK to be slow! > >Well, it sounds like you did a lot more pre-release testing than I gave >you credit for. Sorry about that, and thanks for the work you did. >In a way you're in the same boat as me, in that your test system is a >small site on a fairly slow computer, so it's hard to detect slowdowns >in the code for large collections. I guess I was just venting some >steam over my frustration with the lack of help Geoff and I have had in >maintaining this code. We need a good number of testers who'd be able >to try out new code on large sites, and really hammer away at it to find >problems before we release the code to everyone.
That must be my bad english :-(. You were right the first time. I did not test the pre-release versions. I was just saying I was not sure I would have seen the problem. I understand your frustration but there is also a dilemna about testing this kind of software: you want to be as close as possible as the real case but you certainly don't want to disturb the people using it. As bad as it is sometimes I find no other way than "try" on the live server. And I can't justify doing that with a pre-release version. But I've taken note! I'll try to be more active in supporting free (or GNU) software because I find the support far better than for some paid software. Also the quality is often very good. You should be proud of Ht://Dig. Back when we choose it, we compared it with other software costing thousand dollars and we found it as good. I'm not in touch with search engine technology these days but I think Ht://Dig is still a good choice. Thanks. >The irony in the lack of good solid testing we had in the pre-release is >that the final release ended up going out as pre-release quality, and that >essentially forces all users who are upgrading to become beta testers. >We still get the feedback we need, but it's a high-maintenance way of >getting it, because Geoff and I incur the overhead of putting together >yet another release, and the overhead of dealing with more duplicate >bug reports than we'd get with a more manageably sized group of testers >than the whole ht://Dig community. 3.1.6 was released less than a week >ago, and I've already been alerted to 3 problems in the documentation >(2 of which have been there since the stone age), 4 bugs (1 of which has >always been there, and another which has been reported twice in under >24 hours, i.e. the DEFS problem on SGI), and a potential memory leak or >other performance problem. > >I think the wise course of action at this point would be to wait a >month or so until the bug reports slow to a trickle, then put together >a 3.1.7 release that will be strictly simple, well tested bug fixes and >no new features. > > > I'm glad that "my" patch made it in the new version. I also tought I > > needed a "good case" before bringing a performance problem. > >Well, you certainly did make a good case for this one, and that counts >for something. Thinking back, there was an earlier report about a >slowdown of htsearch in 3.1.6, but the tests in that report suggested >it had more to do with the effects of disk caching than anything else. >I suggested some alternate tests to rule out these effects, but I never >heard back from that person, so I didn't have good reason to believe this >was anything new. I think there may be a lot of potential bugs that >get discovered and reported, but never nailed down, because the only >people who seem to be able to reproduce them don't persevere enough in >their testing to actually isolate them to a good test case that we can >work with. > >-- >Gilles R. Detillieux E-mail: <[EMAIL PROTECTED]> >Spinal Cord Research >Centre WWW: http://www.scrc.umanitoba.ca/~grdetil >Dept. Physiology, U. of Manitoba Phone: (204)789-3766 >Winnipeg, MB R3E 3J7 (Canada) Fax: (204)789-3930 ------- Robert Marchand t�l: 343-6111 poste 5210 DGTIC-SIT e-mail: [EMAIL PROTECTED] Universit� de Montr�al Montr�al, Canada _______________________________________________ htdig-general mailing list <[EMAIL PROTECTED]> To unsubscribe, send a message to <[EMAIL PROTECTED]> with a subject of unsubscribe FAQ: http://htdig.sourceforge.net/FAQ.html
