[Numpy-discussion] Announcing toydist, improving distribution and packaging situation

[David Cournapeau]

On Thu, Dec 31, 2009 at 10:50 AM, Christopher Barker
<Chris.Barker at noaa.gov> wrote:
> David Cournapeau wrote:
>> You really should not try to make the point that Pypi is working for
>> the scipy community:
>
> I think the evidence supports that pypi is useful

Yes - I stressed that it was not working well, for the scipy
community. Not that it was not working at all for python

>, and therefor better
> than nothing -- which in no way means it couldn't be much better.
>
> My personal experience is that I always try:
>
> easy_install whatever
>
> first, and when it works, I'm happy and a bit surprised.

To say that you are happy when it works is telling about your low
expectations, no ? My main point of disagreement with most discussion
on distutils-sig is that I think the lack of robustness is rooted in
the way tools are conceived and used, whereas others think it can be
fixed by adding more features. This, and the refusal of learning how
other communities do it is why I am considering starting with my own
solution,

David