Boost logo

Boost :

Subject: Re: [boost] RFC: Separating Boost.Python from Boost
From: Bjørn Roald (bjorn_at_[hidden])
Date: 2015-05-31 15:21:34


On 31. mai 2015 20:28, Peter Dimov wrote:
> Bjørn Roald wrote:
>> On 31. mai 2015 12:22, Peter Dimov wrote:
>> > Rene Rivera wrote:
>> >
>> >> I've mentioned before that I would very much prefer if we didn't
>> have >> libraries within libraries in the libs structure.
>> >
>> > I've mentioned before that I agree, but will say so again.
>>
>> Is it really a question of libraries within libraries in the boost
>> library sense where a library is maintained by individuals or a group
>> of individuals. Or are this just allowing structure within a library,
>> where subdirectories with its own standard structure may for instance
>> be optional parts of the library which when used add additional
>> dependencies.
>
> It is basically a question of removing the current special case of
> libs/numeric/* and not introducing any others. The libraries in numeric/
> are already in their own separate repositories.

sure, but among other effects such a restriction will have is that
further modularization likely will require a lot of tiny git axillary
repositories to handle optional glue between boost libraries or glue to
optional external technologies. That is if not an alternative, perhaps
superior, method to manage such optional dependencies can be realized.

I sort of dislike the extremely hard tie between a git repository and
the physical layout of directories in it. Some standard structure is
clearly needed, but allowing this to recurse does not seem too hard or
very unreasonable. If someone want to have more structure within their
repository, that ought to be allowed. Detecting subdirectories with the
conventional directory structure is one posibility, another is to
require such subdirs to be declared explicitly when needed, which is
very common in many build systems. Just my 0.05$ anyways.

--
Bjørn

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk