Re: Info menus on text-mode terminals

[Luc Teirlinck]

Miles Bader wrote:

   In the previous discussion I advocated actually modifying and filling the
   buffer text to achieve Kim's layout improvements, instead of using display
   property hacks to do it.  This would allow better layout, and eliminate some
   weird side-effects of using display properties (such as funny cursor
   movement).

Yes, but it would have severe negative effects, as I already pointed
out in the previous discussion.  The current implementation of
Info-hide-note-references and related features is fundamentally flawed
because it tries to reformat Info files as if they were written in a
mark up language.  They are not, they are plain text.  The only way to
ever make such functionality work completely satisfactorily is through
a new option to makeinfo.  (That is probably what we eventually will
have to do, but, unfortunately, currently nobody seems to have the
time to do it.)

One consequence of that is that the refilling as attempted a while ago
did not respect the functionality in (texinfo)Breaks, such as @* and
@w.  I have had to use @w in texinfo, to avoid extremely bad line
breaks.

Remember that texinfo is supposed to be a general purpose mark up
language, not a hyper-specialized language for use in Emacs
documentation.  For some types of text, say mathematical text, being
able to force and avoid line breaks is extremely important.  If the
file contains text of a mathematical nature, line breaks at the wrong
places look disastrously bad.  In general, in any type of more
technical text, if one uses technical symbols that _need_ to stay
together, you _need_ to be able to rely 100% on @w.  If not, texinfo
as a mark up language would be badly broken.  Actually, it would be
more correct to say that Emacs as an Info reader would be badly
broken, but due to the importance of Emacs as an Info reader, that
would necessarily affect the texinfo language as such. 

(We discussed this before.)

   In the context of that discussion, I think you're right, there was
   no mention was made of ttys.  The `tty problem' is even more
   incentive to do it though...

Richard seems to believe that there is some hope to make "space" work
on ttys for integer spacing values.  That would eliminate this bug and
probably eliminate similar, not yet discovered, problems on ttys as
well.  (I doubt that the problem we are looking at is the absolute
only instance where somebody used the display property and forgot to
test his code on a tty.)

Sincerely,

Luc.