> > A DTD describing an "abstract presentation information" (which Nathan
> > suggests) is no more or less than _another_ presentation format. Surely
> > we can aim higher that that.
> > > As usual, my opinions only...
> The trick is to allow device independent presentation while still providing
> the most common semantics. HTML/HTML+ is ultimatly a presentation language
> and there is no way around it. If you cannot convert most DTD's to
> HTML/HTML+ then we are back to square one because it is *impossible* to
> include semantics for everything under the sun.
Nathan was suggesting disposing of semantic information altogether. That
is what I object to. There are clear advantages to identifying semantic
The trouble with saying that HTML/HTML+ is ultimately a presentation language
is that is IMHO a bit short sighted. By adding structure to documents, you
open up a lot more possibilities. Presentation is one use, but why
exclude others by disposing of the semantic markup. Index building, search
in context etc.
> Will we ever be free of the SGML "presentation is evil, evil I say" party
> line? If you want anyone to be able to read your SGML then you have to
> have some presentation somewhere. HTML/HTML+ is a compromise between
> allowing as much semantic content as possible -- while still having a
> finite DTD -- but still allowing a wide variety of data to be encoded
> for (here is that evil word again) presentation.
I agree with you in that we can't expect to have an HTML+ DTD which has
all the semantic elements you would ever want. That should not preclude
trying to come up with a usable set however, with a facility for those
that don't fit.
> HTML/HTML+ is presentation with semantics, keep saying that over and over.
I would rather see it as semantics with presentation :-)
> On the other hand maybe you know something I don't. If you can write a
> DTD that does everything that everyone needs without using any presentation
> then get in contact with Dave Raggett <firstname.lastname@example.org> and see what
> you two can work out. Oh, BTW, most of the actuall presentation
> is in what we are calling a "Style Guide" that is external to the
> HTML/HTML+ document and provides hints to the browser about how to
> render the document (so you can use the authors style guide, or your
> own personal favorite).
The style sheet idea is great. By extracting the information from the
document, you inevitably get more control for less effort. On this
subject, this has been the area of a lot of work in the SGML community -
coding presentation information. Are you familiar with the FOSI
standard (US DoD) or the DSSSL standard. I'm not suggesting adopting
them, but they may contain some useful ideas/pointers to the issues.
Dave and I have exchanged a few mails. I disagreed with the notion of
having the semantic content coded as attributes to <p> and <em>. This
makes the structure of the DTD very flat and I would rather see it more
hierarchical. The <cite> element is a good case in point. I would like
to see this as a container with say <author> <isbn> <publisher> etc as
sub-elements. More hierarchy, more structure, without compromising the
ability to add presentation information in any way.
It also leaves open the evolution of the DTD without the problem of
"Hell - how can I add a content model to an attribute". I.e. if you
code say <head> as an attribute, it can't then contain <title>, <isindex>
> When we talk about presentation in HTML+ we are mostly talking about tables
> and external graphical image layout (e.g., gif's). Someone should make
> a list of the presentation like elements in HTML+.