Re: On PEAR quality issues and natural selection

From: "Paul M Jones" <pmjones@ciaweb.net> > >> This would allow distribution of effort among all users to highlight > >> what packages need attention, and PEAR-QA would have instant feedback > >> on what packages are not compliant. > > > > Another advantage is that people will more easily be able to > > distinguish > > between "good" packages the rest by comparing the ratings. > > Absolutely. At the risk of starting another flame war, the Amazon > model has worked very well to help their customers filter through the > huge volume of choices; a similar model applied to PEAR can lead to > very good things. Dynamic feedback loops and all that. Remember, though, that feedback would have to be bound to releases. For example, it could be that I _love_ MDB but dislike MDB2. Also, I might not like some new features that may be introduced, etc. Just like in Amazon: "This rating refers to a different edition." That would be worthwhile. I really like the feedback and rating idea. In fact, a rating interface was proposed a long time ago to improve QA. At the time of the proposal, of course, it was limited to the QA team. But I'd really like to see a rating system sorta like on CNET or ciao.com (ciao.co.uk). It would make it possible for users to say that for example, the docs are really poor, etc. Some ideas for rating categories: - API - Docs - Stability - Usefulness - Code quality (not CS) Of course, the package-level rating would have to be built from release ratings. The ratings displayed for the package would be a weighted average, weighting released based on the following categories (incomplete list): - Number of downloads - number of users commenting - number and seriousness of bugs fixed in succeeding release Just my 2 cts. Klaus