[Python-Dev] PyPI comments and ratings, really? (original) (raw)

Terry Reedy tjreedy at udel.edu
Thu Nov 12 21:42:37 CET 2009

Guido van Rossum wrote:

On Thu, Nov 12, 2009 at 10:30 AM, Terry Reedy <tjreedy at udel.edu> wrote:

Barry Warsaw wrote:

On Nov 12, 2009, at 8:06 AM, Jesse Noller wrote:

Frankly, I agree with him. As implemented, I and others think this is broken. I've taken the stance of not publishing things to PyPi until A> I find the time to contribute to make it better or B> It changes. That's distressing. For better or worse PyPI is the central repository of 3rd party packages. It should be easy, desirable, fun and socially encouraged to get your packages there. I think his point is that a new book announcement servive is different from a book review and rating service. And that mixing the two is 'socially discouraging'. I do not know what the answer is I would say that publishers disagree -- they seem to really like adding "social" stuff to their book announcement service. See e.g. Amazon (which combines all functions: announcement/promotion, ordering/download, review/comments/rate/popularity).

I use user reviews on both Amazon and Netflix. I notice that both let people rate the reviews (helpful or not), and I sometimes look at those also. Both list items without the say-so of creators, though most will tolerate possible bad reviews in exchange for sales. I very seldom see an item with only one review, so there is usually a mix. There are also ratings averaged across a lot more people.

Part of the pypi problem is a startup problem of initially low numbers. If the only people who bother to log in to rate are the disgruntled, then the ratings/reviews will be biased. I wonder how many of the people promoting the new feature have themselves logged in to systematically rate and possibly comment on every package they have looked at, and thereby kickstart the system with fair responses.

Authors can often respond to magazine/journal reviews in Letters to the Editor. Publishers tend to exercise some editorial control over reviews so as to not make the publication look bad with grossly bad reviews. Even on Amazon, an author could, I presume, add a response to a factually incorrect review.

Terry Jan Reedy

More information about the Python-Dev mailing list