Re: Positioning HTML Elements with Cascading Style Sheets

Chris Lilley wrote:
> > How do you figure that automatic generation of a table of contents, index,
> > or index of figures at an author chosen location changes the semantics of
> > the document?
> 
> I wrote, structure or semantics. In those cases, the change is to the
> structure.

If I put 

<TITLE>This is the Title</TITLE>
<GENERATETOC>
<H1>...
<H1>...

Then the structure *includes* the TOC to me, just as if I put:

<P>For more information on this, please see <GENERATEREFERENCE ID="Section.1">
or <OBJECT SRC="toc.html"> . It is irrelevant whether it exists as a physical 
file somewhere, if it can be predictably rendered.

> > Does automatic generation of list bullets change the semantics?
> 
> No, that would be a style change and is handled by stylesheets.

But in most cases, the list bullet is "rendered" as another textual object on
the client. It is "generated text" that the author expects, in the same way
that a reference or table of contents is.
 
> > How about the text for references?
> 
> Perhaps you could expand on what you mean for that one.

See above.
 
> > Why ship a TOC or list bullet over the web when it can be
> > generated automatically?
> 
> So that indexers can index it and downlevel browsers can see it at all,
> even if not very prettily. 

Indexers can index the document. Since the TOC doesn't include any text that
isn't also in the document, having them indexed separately just seems like a
way of promoting superfluous hits of the type that really annoy me.

Downlevel browsers will have to be taken care of, for a while, but *allowing*
TOC generation on the client does not in any way make it impossible or 
dificult to do it at the server side.

> Besides, we don't ship around list bullets.

Right. Because they can be generated at the client. Which is why we should also
not ship around TOCs.

> We do currently ship around TOC, and these are often machine generated,
> at least in the documents that I write.

So we might as well give authors the option of letting the user agent
do the work since the output of the process is totally predictable and non-
creative like *other style issues*.

 Paul Prescod

Received on Tuesday, 4 February 1997 19:03:50 UTC