Re: HTML 4/XML+CSS2 front-end

[Valeriy E. Ushakov]

On Tue, Dec 29, 1998 at 12:23:00PM -0000, MatЛj Cepl wrote:

> > > are there any thoughts for a lout's front-end for specified beasts?

Ok, after reading your reply I can see that I misunderstood your
initial question.  In the usual terminology frontend is identified by
the source format it accepts and backend is identified by the format
it targets (e.g. gcc has a C++ frontend and a SPARC backend).


> Does it make so much difference whether I would be using HTML4/CSS2
> or XML/XSL?

Not much, but see below.


> So, my idea of the best writing has changed a little bit.  The input
> would be SGML-like (plus some stylesheet -- but when I gave just a
> very short look to DSSSL, I began to panic -- do you really want me
> to learn _such huge_ stuff?)

SGML/DSSSL (without loss of generality, this equally applies to
XML/XSL as well) are targeted to a slightly different usage profile.
It is assumed that you have huge volumes of SGML marked data and you
have a DTD and a DSSSL stylesheet for the DTD written once.  Thus you
invest in the DTD/stylesheet once and you reap benefits afterwards.
Thus it saves you efforts when you're a big company, or if you use
some well established DTD (like DocBook).


>       1) that the whole thing is just the same as many things in open
> source world -- "not yet". It will be really terrible thing when all
> bugs will be taken out of it (footnote-mark in the beginning of line is
> the most prominent example).

As you correctly point out, it's the same as many things on the free
software world.  You could invest in the open source product to make
it up to your task or you could invest in the commercial product that
is up to your task.  In most cases you have to invest and you just
decide which product, free or commercial, to invest into (with all due
respect to RMS, I don't share his hatred to commercial software, but
that's a quite different topic).

Re footnote bug.  First, in case you missed it, I have posted the
simple fix (Lout level, so no need to recompile) for this problem to
the list that you could use while waiting for 3.13 to be released.
Second, every software has bugs and the open source is about being
able to fix them yourself, not waiting for vendor to release the fix.

Vendors are often slow and reluctant to introduce changes because one
man's bug is another man's feature.  Vendor has customers that
invested in their product and fixing a bug may actually break
someone's code, so the fix is put on hold (and I do have actual first
hand experience of this situation).  Open source product gives user an
advantage of customizing product to suit his needs and to fix bugs
that are important for you to have fixed (but again, this means
investment).


>       2) that it has not been widespread as it should be -- which
> means, for example, that I would be waiting for many years for Czech
> spellchecker for it.

Spellcheking Lout is a little bit harder than, say, TeX, because any
word in the input could be a command.  One solution is to create a
cutom dictionary that will list exported symbols as valid words.  The
additional benefit is that spellchecker will catch a typo in a
@RawVeryWideTeggedList as well ;-).  Using it with emacs/flyspell
could be very convenient.


> 
>       3) that some constructs like
> 
>       @TeX &0.01s u 
> 
>       to get just one word "TeXu" (Czech language tends to make
> incredible tricks with every word, even non-Czech), did not seem to me
> quite "logically marked up" :-) I know, it is probably old unresolved
> bug, but well, it is not quite what I expected.

Since Russian is an inflected language as well, I see your point.
However, since declination of words written in latin script usually
requires an apostrophe to sepatate the stem form the ending, @TeX'u
(with Russian `u') will do the trick for me.  In your case you could
use @TeX{u}, perhaps.


> You are saying, that HTML4 does not have a future. 

Huh?  I haven't said this.


> Well, probably you are true, but tell me, where should I find any
> free software supporting XML/XSL now. OK, so I shall stay with
> HTML4/CSS for awhile, which should be convertible to XML easily.

A central catalog for all SGML related material is
http://www.oasis-open.org/cover/


> Therefore, my dream has changed again. It should really tiny program
> (I guess not bigger than lout) able to transform valid HTML4/CSS2
> (with all things necessary for printing like "page-break-after:
> avoid") document into PostScript. I do not see any reason, why such
> thing should be impossible, but unfortunately I do not see anybody
> able to help me with it.

Everything is just a "simple matter of programming" ;-).  Seriously,
parsing in-the-wild HTML is a challenging task since people constantly
*misuse* it.  I bet that 95% of the web will not pass through an SGML
parser.  So web browsers actually parse *something* that resembles
HTML (as SGML application) but has so vague semantic that they have to
handle lots of special cases.

If you stick to the rigid valid HTML, that using XML/XSL is just one
little step away (SGML/DSSSL being the next step), but making this
step gives one (in theory, at least) significant benefits, as you step
on the firm ground of *standard*.


> The one possible solution to this problem (not the best) would if
> there is HTML4/XML front-end for Lout. Therefore I asked you,
> whether you know about anything like this.

If you are looking for an open source SGML/DSSSL engine, than Jade,
the monumental effort by James Clark, is definetly the first candidate
to consider (actualy, it's Jade that I had in mind when I wrote about
Lout backend for a DSSSL engine).

There's a number of non-DSSSL SGML formatters as well.  In particular
documentation project for a Debian GNU/Linux uses a simple DTD that is
converted to HTML and PostScript (via Lout).



> have a nice day and sorry for so terribly long message

You are welcome.

I think that this boils down to a question: "where to invest".  Some
open source projects are mature enough so that you usually can "just
use it" because other people has already invested much in these
projects and you can just enjoy the fruits (e.g. gcc).  But SGML only
gaining "momentum" now under XML disguise, so there's no esablished,
ready to use open source tools yet, though Jade will be probably in
this position within a couple of years.

All this being my random thoughts, anyway...

SY, Uwe
-- 
address@hidden                         |       Zu Grunde kommen
http://www.ptc.spbu.ru/~uwe/            |       Ist zu Grunde gehen