Using CPAN Ratings as a bug report
CPAN Ratings are a great idea. Unfortunately it seems that in some cases they are either unused (in places where they should be) or misused (in places where they shouldn't be used).
While some distributions (which are very recommended to use by the Perl community) do not enjoy having ratings at all, I've noticed some people putting their personal gripes as ratings.
Theoretically you should be able to do that, true, but these ratings don't act the way you might assume they would. If you have a dist and I rate your first version as one star (for lack of tests, or because I do not see it useful for me), you have a single star (out of five) marked until someone changes that. I've seen ratings that existed since the first development version of a distribution (which usually say "this sucks, how do I use it?!") - leaving me to believe that they cannot be deleted, perhaps only to the original poster who might be MIA.
If you read ratings every now and then, you might encounter people who rated a five star just to negate a single star that was given way back.
The ratings that bother me the most are those that reflect things that can go in a bug report. Missing docs? file a bug report. Missing feature? file a bug report. Missing tests? file a bug report.
The problem with the ratings is that they tarnish the name of any dist as something that is permanent (I don't remember ratings that were deleted so I don't know if it's possible - I might be wrong on this) with something that is very temporary.
If someone opened a very low rating because of documentation and I add it, should I now fight it off with marking the five star review? Should I contact the author of the tarnishing review? What if he feels the documentation is not adequate enough? Does it still deserve a single star? Can we settle for three? What about three and a half? It's absurd.
Please stop adding ratings that can otherwise be reported as bugs/feature requests and open tickets instead. Some of us even use more than one ticket system, answer to emails and available on IRC.
P.S.:
If anyone knows whether they can be deleted and how, there are some (well, at least one) that I would personally like to delete and resubmit as a bug report instead.
Gots to do the bloggingz
I've wanted a more "Temporal" style of ratings system, simply because this star rating thing can work the other way.
ie: Somebody could think Switch.pm is just awesome 10 years ago, and they're no longer using Perl, and the community at large is saying "Hey, don't use this", but the 5-star rating would remain until somebody added 1-star ratings to compensate.
Ideally, the current star rating would be based on some function of time smoothing to give a representation relative to *now* instead of 10 years ago.
Myself, I'd like to see graphs of ratings/popularity vs time of varying modules, and have a table of risingly popular and risingly unpopular modules.
That is an excellent idea!
it is
Not to mention people misusing CPAN Ratings for discussion (replying to another user's review, or another reply, and so on).
I agree wholeheartedly that some of the review belong in the RT. But you know what, I think the biggest problem for CPAN Ratings at the moment is that... there are not enough reviews :) When I browse search.cpan.org the majority of distributions do not have reviews at all, I'm wondering whether most of the distributions are only used by the author (probably true).
Also, at least CPAN Ratings displays the version of the module reviewed, the date, and the "x out of y people find this review useful" thing. Discerning readers can at least take these into account when reading a review. And, interested parties (the author included, as is often the case) can always add another review to respond to the original outdated one.
It's funny because I got one of this reviews just yesterday, for a module at version 0.01 and clearly marked as experimental on the docs. I hate it!
Anyway, I agree with Steven Haryanto, the problem with CPAN ratings is that there are very few reviews.