Re: perl customization interface stability

[Karl Berry]

Hi Patrice,

    get_conf function there is an issue, because it is different when used
    from the main program (ie from texi2any) and from a converter, 

Ah.

    be forced to use a suboptimal API, 

Sure, I understand and agree, in principle.

Of course, API changes will always be possible.  But at some point it
should be possible to say that changes will be additions rather than
backward-incompatibilities, at least in most cases.  Maybe get_conf
could already fall into that category?  

As far as user expectations go, if we put something in the manual,
people will use it, regardless of how we label it.  They don't have any
rational alternative.  So if you don't want to commit to supporting
get_conf/gdt/set_from_init_file/whatever, then I think we should just
take them out of the manual, as we already did with all that other
"experimental" stuff.  I don't see a difference.

Also, FWIW, since customization is (supposed to be) one of the big
advantages of the new implementation, I think defining/documenting an
API you are comfortable with advertising to users should be a very high
priority for 5.1.  Maybe the highest.  The sooner that happens, the
sooner people will be able to use it.  Right now, in the absence of
anything like that, people will just poke around in the Perl sources
and do whatever works and/or take whatever you send them in email.  All
of which could well lead to hassle in the future when the "internals"
they end up using change ...

Thanks,
karl