>>>>> On Fri, 18 Dec 2009 23:46:29 +0200, Serguei Trouchelle >>>>> <[email protected]> said:
> Most valuable tests (IMO) are FAIL tests, because they indicate that > something's not right and raise author's attention. We had about > 30,000-40,000 FAIL tests monthly during last 6 months, and it's less > than 100,000. > It may be a good idea to send FAIL/UNKNOWN reports and hold on with > OK/NA. I disagree. I'd start from the other end: the least interesting tests are those that have been reported several hundred times. If everybody sends only fails we lose the ability to analyse the reasons for the fails. Once in a while every distro gets a fail and after a while the whole CPAN would look bad and we would not be able to tell if it is true. If you're able to limit your smpokers to uploads within the last 6 months you're most likely throttling well. And before somebody argues that even old uploads may be interesting: I wholeheartedly agree. It's a pity if we have to throttle and I believe we do the least harm by concentrating on recent uploads. -- andreas
