Re: bookmarks for concept index

[Werner LEMBERG]

> Seriously: I'm a bit worried about introducing new language features
> matter-of-factly.  Texinfo doesn't have a graceful backward
> compatibility feature, so a manual that uses a newly introduced
> directive _forces_ users to upgrade to the latest Texinfo.  That is
> not nice, IMHO.  Most other tools I know and use don't behave like
> that, they always try to make changes back-compatible.

I think you are too conservative here.  Usually, a new version
introduces new keywords.  This has happened with TeX, this happens
with autoconf, python, etc.  IMHO, the texinfo language isn't carved
into stone since it still has a *lot* of deficiencies, making it
really suitable for C and Lisp only.  C++ already gives some
difficulties, and documenting the groff language is really hard, as
you may already know.  Well, I can avoid problems by simpling omitting
some of the best features of texinfo, but I don't want that.  I want
texinfo a general purpose documenting system, and until this isn't
reached (to a certain degree), I won't stop making suggestions for new
keywords if the current ones don't suffice.  My address@hidden' suggestion,
for example, which I still believe is a good idea (I'll try to provide
a patch if time permits, except you say that you dislike it completely
-- BTW, you haven't commented on my last reply regarding this topic).

> So I'd much prefer changes we make in the language are
> back-compatible, in the sense that they do not cause older versions
> of makeinfo and texinfo.tex to choke, or crash and burn.

This is unavoidable to a certain degree.  But usually
`backwards-compatible' means the opposite, i.e., old documents can be
run without problems using newer versions of texinfo.

AFAIK, it is common practice (or even in the GNU guidelines?) to
provide processed texinfo files with a distribution -- this avoids the
version problem completely.  Consequently, any changes to the format
of info files must be done with great care, but changes to the texinfo
language are not affected.

> Perhaps we should also add a feature which ignores
> unknown/unsupported directives.  That won't help users of older
> versions now, but it will help in the long run, when future versions
> introduce incompatible changes.

A good idea.  What about a switch which makes unknown keywords and
environments a warning instead of an error?  I know Eli will hate me
for this, but I suggest a @version keyword (similar to LaTeX) so that
makeinfo/texinfo.tex aborts gracefully if unknown features are used.


    Werner