Re: [gentoo-dev] Re: [RFC] Masterplan for solving LINGUAS problems

[Michał Górny]

Dnia 1 czerwca 2016 16:03:40 CEST, Mart Raudsepp <leio@g.o> napisał(a):
>Ühel kenal päeval, K, 01.06.2016 kell 15:19, kirjutas Michał Górny:
>> As for LINGUAS, it should be left as a toy for advanced users and not
>> presented as a recommended solution.
>
>There is nothing advanced in it for the user, only the mess we have
>created with package manager behaviour and mis-use of it (the order
>matters case; which I believe is long eradicated).
>We are a source based distribution, and gettext/intltool upstream
>LINGUAS behaviour is perfect advantage for our main use case of
>customizing ones own system and almost always building things from
>source, only using binary packages before an upgrade as a backup, if at
>all.
>So it's natural to use the way that really build only the support you
>want. This is what LINGUAS gives you, when the PM doesn't happen to
>munge it.
>
>Hiding this away under some toy for advanced users is not in our spirit
>of Gentoo, as far as I would judge.

You forget the important point that it's done silently and implicitly, with no clear way of knowing which localizations were actually discarded afterwards.

And the fact that currently LINGUAS affects both packages listing the flags and not doing so is causing even more confusion.

>
>But this is a matter of documentation at this point, in principle I
>agree that SRC_URI extra downloads should be under a different naming.
>
>INSTALL_MASK groups for locales is what I would consider a convenience
>for binary package builders in a wide environment where language choice
>to the end user preferably gets filtered on deployment in a site- or
>machine-specific manner. Or a toy for advanced binary distribution
>creators, if you will. A way for binary packages to provide almost as
>good support for LINGUAS as source packages (but not quite).
>That said, supporting our binary package ecosystem is very important,
>and I applaud these efforts. The proposed INSTALL_MASK improvements are
>very useful for many other cases as well. For source-based users as
>well (openrc init scripts, systemd unit files, gtk-doc documentation,
>etc)
>
>Either way, the masterplan works out, I just don't think we need to
>wait for INSTALL_MASK groups here in any way. The reminder is a matter
>of documentation, a matter of perspective.
>This l10n.eclass PLOCALES nonsense needs to go ASAP.
>
>
>Mart


-- 
Best regards,
Michał Górny (by phone)