On Tue, Aug 17, 2010 at 4:21 PM, Bram <q...@perl.wizbit.be> wrote: > NOTE: I have not actively followed the cpan-testers-discuss mailing list nor > the switch to CT2 so what I say may have been said before. (appologies if > that is the case) > > The way I see it there are (at least) two types of reports: > > A) those that are coming from cpan smokes/automated testing: > My guess is that most of the email reports are comping from these systems. > (Are there any stats on this?)
I suspect that over 95% of the reports come from automated smoke testers. Most of those have now converted to CT2.0. > B) those that come from users when they install a module on a system: > > My guess is that this is a limited amount of email traffic and 'we' can not > be expected that everyone knows the details of the reports and the move to > CT2. The email auto-responder from perl.org is only being turned on now to notify those who haven't yet heard the news (which has been blogged about extensively, flagged on perlbuzz, etc.) > Assuming the emailaddress is switched off completly at some point: what I > expect to happen then is that a (large?) part of the reports that are send > by users will be lost because these users are unaware of the need to upgrade > the test reporter module and/or can't be bothered enough to upgrade it. Only a small number of reports would be lost in absolute numbers, but I think it would be a terrible loss because these reports represent "real world" systems with a wider variety of existing installed modules, etc. If you wondered why I'm checking email and trying to be helpful while on vacation this week, it's because I do value the small, individual tester's reports so much. > Is it an option to keep the emailaddress but keep the quantity send to it > low (read: no automated smokers) so that it does not have a big impact on > the perl.org infrastructure? It's not an option, for several reasons, though it was considered. For one, it's not just the email traffic but the back-end storage that was the problem for perl.org. (Solvable problem, but not where they wanted to be spending *their* volunteer admin time.) Another issue is that cpan-test...@perl.org takes reports from anyone without prior subscription to the list. That makes it a spam magnet with more overhead to do textual analysis for filtering rather than filtering just on sender address. For another reason, the medium term goal is to have reports consist of structured data, not "email text" reports. So even if reports could still be sent by email, the "plumbing" would all change anyway. I expect there will be a period where individual testers get really annoyed at Barbie, me and others involved in the project and may drop out rather than bear with us through the conversion. I really hope that is the exception and not the rule and I hope they come back once we make it easy again. I ask everyone to be patient. If the documentation isn't clear, ask questions. If you get helpful answers (or even if you don't), please update wiki.cpantesters.org to make it less confusing for the next person. If you think you can write code to smooth the transition, almost all of the source is available on my github account ( http://github.com/dagolden/ ) -- David
