I'm now settled in my new job (and new appartment), the new and
improved CPANTS is running on a new server (provided by yi.org, thanks
again to Tyler MacDonald!). So basically all the time I can spend on
CPANTS will go into new tests (eg a check if used modules (minus stuff
in Module::CoreList) matches PREREQ_PM).
That would be really nice, and thanks.
In fact, you have my continual thanks for making something that means
your can actually keep an eye out for problems across acres of modules.
I think pretty much everyone with 20+ modules needs something like it.
- Until I grok PPI and merry it with CPANTS, testing distribution
kwalitee is basically the only halfway serious option. Even this
doesn't work all the time (see has_test_pod*).
Dist tests are low-hanging fruits. But I'll promise I'll reach
further. Later...
You can play a little with PPI if you wish, but it still can't run _all_
of CPAN without overloading memory usage from various enormous Perl files.
That said, I'd like to see a metric called no_insanely_large_files to
trap everyone with a single file larger than about 250k. Almost all of
them are trouble and could probably use various techniques to shrink
them down (seperating the docs into *.pod or something)
I do promise to get back to you once PPI is safe for all of CPAN without
going haywire.
- CPANTS as a multiplayer online game is an easy way to get peoples
attention without totaly offending them. I /could/ send an email to
everybody on CPAN with some 'helpfull hints' on how to improve
kwalitee. I guess the biggest effect would be to get added to some SPAM
blacklists etc...
But with the tongue-in-cheek 'highscore lists', people get
interested/hooked and DO improve their code. I got several mails of
people who discovered semi-serious problems in their code (eg missing
'use strict' statements) because they checked their CPANTS ratings.
If people want to 'cheat', that's ok for me. As soon as I have some
time to spend on the issue, I can improve the tests (but that's rather
low on my todo list, as I like to assume that we are all grown-ups and
do not need faked cpants ratings to boost our ego (I might be
wrong...)).
And no, I won't take the fun out of CPANTS.
I actually completely agree with you here.
If you are going to score something, people will inevitably want to
compete. And if they compete they will inevitably also cheat.
So there's only two choices.
1. Make it so easy and obvious how to cheat there's just no challenge.
2. Start a cold war (ala page rank) with the authors
And I doubt you either have the time or the inclination to do 2.
Simply adding additional tests over time will deal with most of the
problems of gamesmanship, because people won't have to go to such
extreme lengths in order to do well.
You just need to be aware that people will naturally game things, and so
adding metrics where the gaming is a positive effect (or at least, not
negative) provides the best result.
And remind me to talk about taint_safe_versions at some point :)
Adam K
