[Catalog-sig] Package comments

[Ian Bicking]

On Wed, Nov 4, 2009 at 3:55 AM, M.-A. Lemburg <mal at egenix.com> wrote:
> If PyPI is meant to go social rather than being purely a search
> engine, then the ones who put their content up on PyPI should at
> the very least be allowed to reply to comments and to be fair,
> ratings should be visible to all and only be possible together with
> a comment.

For low ratings (e.g., 1 or 2) forcing a comment might be interesting.
 Higher ratings (4-5) are usually not that interesting ("it works as
documented").

> Regarding the usefulness of such a feature, take the PIL package
> as example:
>
> http://pypi.python.org/pypi/PIL/1.1.6
>
> """
>  Package rating (3 votes): 4.66666666667
>
>    * 4 points: 1 vote
>    * 5 points: 2 votes
>
> Ratings range from 0 to 5 (best).
> Package Comments:
>
>    * Hugely stable, been around forever, and works. Unfortunate distribution naming causes problems
> with setuptools. (chrisw, 2009-10-05, points)
>    * super cool lib - #1 pick! just sad to see the packaging difficulties. (jensens, 2009-10-05,
> points)
> """
>
> Those comments are not really all that useful for a user,
> since they put too much emphasis on a non-package related issue.
> This is like saying: "Great bag, but doesn't come in pink, so
> only 4 points.". An educated user will notice, a casual user
> will just see the negative vibes and not bother with PIL,
> since it "causes problems" - now *that* is sad.

To make it more useful I just added a comment to point to a
easy_installable version of PIL.  In my mind this basically resolves
the packaging issue, through sideways communication among users of
PIL.  This sideways communication is useful.  It just occurred to me
that virtualenv-distribute (a fork of my virtualenv package) had
functionality now duplicated by the latest virtualenv release, so I
left a comment to that effect on that package (I also had to rate,
which was annoying -- I just wanted to add information).  I could have
contacted the author, and I guess I did by leaving a comment, but this
way I can contact the author and everyone else at the same time (like
emailing a mailing list instead of personal email).

So... I guess after thinking about it, I'm seeing more potential for
comments.  That said, I also think it is entirely reasonable to give
package authors dictatorial control over comments, allowing them to
delete anything.  I don't believe that power is any more likely to be
abused than comments themselves will be abused.  If deletion keeps the
comments in the database (but hidden) and is required to be
accompanied with a justification (just free text, with maybe a
shortcut for spam) then if there really are problems then they can be
talked out.

I also think we can try to resolve some problems through the use of
human-readable text (on the site and in the emails the site generates)
rather than through code.  (Oops, looking back I was supposed to
provide some text for the comments, but I haven't...)  That is: if we
agree some things are inappropriate for comments, and some things are
particularly useful for comments, then we say that explicitly on the
site.  I expect commenters will pay attention to those guidelines.
Same thing for comment deletion.

-- 
Ian Bicking  |  http://blog.ianbicking.org  |  http://topplabs.org/civichacker