[Distutils] Removing the aspirational aspects of PEP 426

[Nick Coghlan]

On 29 October 2015 at 02:09, Robert Collins <robertc at robertcollins.net> wrote:
> On 28 October 2015 at 23:39, Nathaniel Smith <njs at pobox.com> wrote:
>> On Oct 28, 2015 3:25 AM, "Nick Coghlan" <ncoghlan at gmail.com> wrote:
>>>
>> [...]
>>> From an sdist metadata perspective, though, I think the right thing to do
>>> is to descope PEP 426 to just the stuff we *need* for the build system
>>> improvements, and defer everything else (e.g. JSON-LD, SPDX, richer
>>> dependency semantics, etc) to a future metadata 3.0 proposal (or potentially
>>> metadata extensions, or 2.x format updates).
>>
>> I think PEP 426 is actually orthogonal to these proposals. AFAICT, the only
>> reason Robert's PEP as written requires PEP 426 is that he needs a standard
>> serializable format to list dependencies... but he actually defines such a
>> format about 10 lines above for the static bootstrap-requirements key, i.e.
>> a list of specifier strings. So it actually makes more sense to use that for
>> dynamic requirements too for internal consistency, and leave PEP 426 out of
>> it.
>
> pip requires:
>  - distribution name
>  - install_requires [+extras]
>
> today. It will want external dependencies in future (and in the spec I
> put forward build dependencies would be obtained earlier so could be
> skipped).
>
> I'd rather not invent a *new* format for handling both of these, but
> i'm ok if Donald and Nick specifically are.

They're probably a good candidate for the version specifier and
environment marker treatment: extracting a "dependency specifier" PEP
to give us a building block we can use in the higher level specs.

Cheers,
Nick.

>
> -Rob
> --
> Robert Collins <rbtcollins at hp.com>
> Distinguished Technologist
> HP Converged Cloud



-- 
Nick Coghlan   |   ncoghlan at gmail.com   |   Brisbane, Australia