Re: Netscape 1.1 and Accept: headers

   Date: Mon, 6 Mar 1995 12:29:42 -0800 (PST)
   From: Brian Behlendorf <brian@wired.com>

   Comparing the consequences between a almost-HTML-3.0 browser viewing 
   HTML 3.0 pages and a not-more-than-HTML-2.0 browser viewing HTML 3.0 pages, 
   I think the former case is much preferable, particularly if it helps 
   encourage the almost-HTML-3.0 browser become fully-HTML-3.0.  

	   Brian-dash-dot-dot-dash-Behlendorf

That's a matter of taste, but when I point Netscape 1.1 at the W3O's
HTML3 tour (http://www.w3.org/hypertext/WWW/Arena/tour/start) I do see
real problems.

One fairly serious bug that shows up immediately is that the
inline images in <fig>s don't show up, leaving the caption text
floating well out of context:

  As you can see, HTML 3.0 supports text flow around floating figures.
  Other major additions include fill-out forms, tables and mathematical
  equations, and features for greater control of layout. A pause after
  a hard spell of fishing    ) Time-Life 1987

This problem is, in a sense, worse than the difficulties with tables
--- when a table-impaired browser tries to deal with a table-laden
page, you can at least see that there's stuff there which is being
handled improperly.  If a page uses <fig> and the browser doesn't know
about it, it may not be immediately obvious what's going wrong.

(One way around this would be to just not use <fig> until you can have
some confidence that a browser that claims to accept HTML/3.0 really
means it --- but that could delay the adoption of <fig> in cases where
it might actually be useful).

Another, lesser problem is that authors will expect the graphics
associated with a <note role="blah"> ... </note> to provide context
for the enclosed text.  So, when pages using those tags show up in a
browser that doesn't understand them, that text is likely to be a
little hard to read.

Also, <math> shows up as a garbled version of the source, which is
sometimes decipherable and sometimes not (when an unrecognized <over>
tag has vanished, the structural information about the formula that it
conveys is irretrievably gone along with it).

If the only way to fix the table problems is to break all these other
things, that still might be the right thing to do, but collectively, I
do think they add up to a headache which we should avoid if there's an
easy way to avoid it.  Putting tables into HTML/2.1 or 2.2 and doing
type arbitration based on those version numbers would avoid it.

rst

Received on Monday, 6 March 1995 18:01:01 UTC