Re: spec for @inlinehtml etc.

[Karl Berry]

Hi Patrice,

    If I am not wrong this would mean that the main difference with
    normal Texinfo code would be that text would be raw text, that is,
    no escaping of special characters.

It's true that one of the original ideas of the raw blocks was to allow
use of bare { and } without causing errors.  Important for TeX, although
braces are not a big part of html|xml|docbook.

But I think most commands should not work.

    certainly be to consider those blocks as preformatted (as if in
    @example and similar) and in addition mark text with type 'raw'.

Thus making the output not be typewriter by default?  (Which should
surely be the case for raw blocks.)

    Also no paragraph?

I think there should be no automatic paragraph generation.
@html
x

y
@end html

should *not* autogenerate
x

<p>y

If the user wants a <p>, they should put it in themselves.  That's what
raw mode is all about.

    I think it is better to do the right thing for such an important issue
    even if it delays a bit the release.

Yes.  Though I find myself puzzled as to the best thing to do.  The
current behavior of tp is completely consistent and logical, in its way.
But since it is so far from the behavior of @tex, I'm not sure it's best
for users.  On the third hand, I'm not sure if there's any real use of
the raw blocks except for @tex.  I don't recall ever coming across them.

With TeX, what happens with Texinfo commands is not exactly specified --
if something works, then it works.  If it doesn't, then it doesn't.  I
haven't tried it, but I feel sure that something like @anchor or @node
or @menu would fail.  I doubt anything dependent on newline-parsing
(@table, @example) would be reliable.

My hunch is that the most useful Texinfo commands in raw blocks are the
simple character-producing commands, like @@ -- the "insertions within a
paragraph" section of the refcard, to a first approximation.  Exception:
@math{} seems fraught with problems with me.

The simple markup commands ("marking words and phrases" in the refcard)
also seem viable, except for @verb, which is nonsensical in raw
contexts.

@errormsg, @c(omment), why not.

Cross-references ... hmm.  It seems harmless enough to allow them,
though probably not useful.

All else seems doubtful to me.  Certainly nothing involving conditionals
or document structure makes sense.

Wdyt?  Perhaps we should ask for input on bug-texinfo, at least if
anyone has ever used @html, etc.

Thanks,
k