[Catalog-sig] Fwd: [ pypi-Bugs-3488989 ] Need a way to list all package versions
techtonik at gmail.com
Wed Feb 29 16:57:22 CET 2012
On Tue, Feb 28, 2012 at 2:35 PM, Richard Jones <richard at python.org> wrote:
> On 28 February 2012 22:07, anatoly techtonik <techtonik at gmail.com> wrote:
>> On Sun, Feb 19, 2012 at 12:07 AM, Donald Stufft <donald.stufft at gmail.com> wrote:
>>> On Saturday, February 18, 2012 at 4:01 PM, anatoly techtonik wrote:
>>> 1. does anybody else also thinks that having a way to list all package
>>> versions available on PyPI would be awesome?
>>> It's possible via the API to list all packages regardless of hidden or not.
>>> The Web UI doesn't support it. I'm assuming from the ticket PyPI is happy
>>> with the way it works, but fwiw you can see all the package versions for
>>> Sphinx at http://crate.io/packages/Sphinx/ (click the All Versions tab).
>> Thanks for replies about API, I used XML-RPC myself, but it's not a
>> good user experience to require Python shell to get information from
>> the web site. http://pypi.python.org/packages/source/S/Sphinx/ really
>> helps to see what packages are available using browser alone, but
>> unfortunately it is a hidden magic.
> When PyPI was created there was a design discussion and decision made
> that the site implement the most common case which is that package
> authors expose only one version, the most recent one. Unsupported
> versions are deliberately difficult to find. That's the point. The
> intention is to reduce the authors' maintenance burden by discouraging
> people from using older versions of their software.
> Your example project, Sphinx, advertises precisely one current
> version, 1.1.2. This is the one on their website and it's the one on
> Some packages expose more versions because they have multiple release
> branches and that's their choice. More versions are findable on the
> site for those packages.
> And, as has been pointed out, for data mining applications you can get
> access to the full data.
Thanks for the proper explanation. At last, after 11 days and
catalog-sig subscription. =) This proves that SF tracker is useless
and fails to communicate with users. It would be interesting to
glimpse at the archives of this design discussion. I can not agree
that users should be deliberately limited.
More information about the Catalog-SIG