Re: bookmarks for concept index

[Akim Demaille]

>>>>> "Karl" == Karl Berry <address@hidden> writes:

Karl>     Harder: they can't just simply AC_PREREQ the version they
Karl> use: they I don't think it would be a good idea for every
Karl> Texinfo document to start AC_PREREQ'ing their version.  Most
Karl> documents don't use any new features and will work fine with any
Karl> version of Texinfo.  Yet if we make it easy, authors will put
Karl> AC_PREREQ(4.2) or whatever and thus cause a gratuitious failure.

I have another reading of this: it helps pushing old versions away,
and I really think that's a major advantage.  I perfectly understand
how bad my position is for some people, but really, we are wasting a
huge amount of time maintaining backward compatibility with things
that simply should no longer exist.

This is Free Software, and a reasonable amount of backward
compatibility is needed, but not more.

Let me take texi2dvi as an example: because of the (*£&$(*£ backward
compatibility there are crucial options, such as --batch, that
Automake cannot use because we don't know if it is a texi2dvi recent
enough.  It results in poorer code.  The same applied with old version
of Autoconf and so forth.

Karl> I'm not yet convinced there is anything we can do that will be
Karl> all that much better than the way it is now, where new commands
Karl> simply cause an error message.

Agreed.  But consider the fact that an `upgrade your Texinfo' message
is not something you will hear about.  An error on address@hidden'
is something you can receive as a bug report, or maybe the author of
the Texinfo file.  In that case, time is wasted by (i) the user who
sent a bug report, (ii) Texinfo maintainers, and (iii) the package
maintainer.  And it will end in the user installing a newer Texinfo
anyway.