Re: Formalising and documenting the current locale maintenance regime?

[Roland McGrath <roland at redhat dot com>]

> I believe it is a good idea to formalise and document the current
> locale maintenance regime.  Do you agree?

If it makes it easier to maintain them all, without discouraging people
from contributing, then sure.

> At the moment, there is a set of people associated with each locale.
> It is the original author, and the people contributing to the locale.
> It is always a good idea to get these people to approve all change
> requests, before asking for the change to be done in the glibc CVS.

Yes, but primarily because those tend to be the people with access to the
relevant references and the ability to easily test any potential change for
practical issues.

> All change requests should be backed by some reference documenting
> that the change is correct.  I still do not understand the range of
> references accepted, but official standards from some language council
> seem to be well received.

I don't think we have 100% well-specified criteria.  In many countries, the
government has official bodies that issue documents on such things.  If
such bodies exists, then only current government specifications are fully
sufficient.  If there isn't a clear government specification for it, then
it's less clear-cut what is appropriate to require.

> New locales are accepted at face value, and the submitter is assumed
> to know what he is doing.

More or less, though hopefully they come with references to appropriate
government documents from their country to back up their data.  If we had a
person prepared to do the work, it would be ideal to have someone review
the new locales and check the references before they go in.  Would you like
to volunteer?  It might also be reasonable to institute a policy of
notifying the GNU translation team covering the language in question and
requesting that interested people from the locale review the data and cite
any problems with references of their own as well.

> All new locales must use language and country codes from ISO 639-[12] and
> ISO 3166.  

Yes.

> If a submitter is unable to convince the ISO standard maintainers that
> the language or country exist, the glibc maintainers are not interested.

I wouldn't say that.  So far it has not come up.  If there is a case where
the ISO standards maintainers, once contacted, steadfastly refuse to
acknowledge the existence of a bona fide human culture or language, then we
will address the case on its own particulars.

> I suggest adding such information to some document in glibc, and
> making a list of the people associated with each locale.  This list of
> people is the "sub-maintainers" of the given locale, and will be
> responsible for forwarding change requests to the glibc maintainers.

That would be a fine thing.  manual/maint.texi would be the place to add a
new section about how to go about adding locales to glibc or maintaining
the existing ones; send one in and we will review it.  To add a list of
locale maintainers, we need some particular person to be responsible for
keeping the file up to date and making sure that the people named in it can
in fact be contacted to address issues in their locales.  (I am presuming
you want to volunteer.)  Then the thing to would be to add a file
localedata/MAINTAINERS containing the list.



Thanks,
Roland