[Catalog-sig] Package comments
renesd at gmail.com
Wed Nov 4 11:41:41 CET 2009
here's some ideas and observations from other places rating python projects.
Depending on the community(and particular packages), I'm not sure if
it will attract trolls or not. As an example, the comments on the
pygame website I can't remember negative comments on there. I'm sure
there have been some though. Hopefully pypi comments will be just as
Feedback is good under any circumstance... even bug reports :) Some
people don't have bug trackers, or forums... and never get any
feedback for their packages or programs.
Ratings like used in pyweek are also useful... but that has various
different aspects for ratings, not just one number. The games are
rated under different categories:
* Production (graphics, sound, polish)
In the past there were other categories (like 7 of them). Things like
sound/music, comedy etc.
An important rating was 'Did not work for me', or 'Not Applicable'
(NA). This allows people to quickly tell the author something did not
work for them.
A few suggestions for rating categories...
As you can see, which categories you use for ratings can be important
for different types of software, and will also move some people
towards improving their packages in those areas. A package may have a
rating of 10 for tests, but 0 for documentation. In a 0-5 rating
system, this information is not available to the person. This is why
some really awesome things in 0-5 rating systems get rated 2.5... if
50% of people think it's a 5, and 50% of people think it is a 0.
Even with a wider range of categories, a project could still be
Awesome(tm) but only get ratings of 0 in all the categories.
Some people just want to upload code, and do not want to be
critised(constructively or not), or reviewed. Should we be caring
about what these people want? I think so. Critisism does take a long
time to get used to... and for some it is a way of life, but others do
not take it very well. Perhaps this is just a thing they need to get
used to. Really, people should just be able to share their code
without having negative things happen to them. We want people to have
fun with coding python right? This is why I not-at-all-strongly-agree
that authors should be able to opt out/in of ratings, and reviews if
they want to - even though I personally like them.
Ohloh is another website which rates open source(including python) projects...
Ohloh only has a 1-5 rating, and also offers a way to review projects.
Rather than just comments, they have specific 'reviews'... so it
tends to mostly get reviews, not free form comments. Reviews also
have a well understood form and have an academic history. You can
tell a good review from bad easily if you have the academic
background. Also reviews take the form of 'wow, this is awesome... I
wish it could do X though...', which are just as valid and useful as
other types of reviews. Even short reviews like 'I tried this
package, but I like X package better' are useful.
Ohloh also provides a number of other project metrics which they (try)
to cover. You can look at things like activity, code size, number of
contributors, and things like that. Much like the previous code
'kwalitee' project which used various automated metrics on pypi
packages. Where ohloh tries to rate packages against other ones, I
don't think that is necessary. It's done to create competition, and
controversy... which aren't necessarily things people want. Having
top 10 rating lists, and such are not needed really... ratings can be
done on a per project basis and still be useful. They use other
things which are not really fair... like number of commits, number of
lines changed and other such things to rate contributors to a project.
These can be useful, but they do not give credit to many other people
who contribute to projects (eg, people applying patches, people
writing docs and tutorials, and the list goes on).
Free-form comments can be good to allow the community to use them as
they wish. Allowing users and authors to communicate with each other.
Communication is a multiple way process... if authors can not reply
back, it is not communication. However it is still communication
between the different users of packages.
In conclusion, my rating is...
( ) Bad
More information about the Catalog-SIG