RE: CSS1 and tables

Hmm.  Well, I suppose I am somewhat personally to blame for
inconsistencies between IE 3.0 and IE 4.0 with respect to CSS support; I
apologize for the inconveniences it causes the designer community
(believe me, I DO hear about it).  However, in my defense, I'll offer
the justifications for the most major of these inconsistencies-

1) It needs to borne in mind that IE 3.0 (final) shipped approximately
six months before CSS level 1 was approved as a W3C Recommendation.
Much work was done in this time period - mostly fleshing out, not direct
changing (although some fairly major syntax changes, like font-weight
values and the addition of sub-properties, did occur), but much of the
fleshing out caused "oh, that's what was meant"-style forehead-slapping.
IE 4.0, for the most part, attempts to adhere to the letter of the CSS1
recommendation - and therefore, breaks with IE 3.0 in a number of ways.

2) IE 3.0 did not correctly (even for the time it was implemented)
implement a couple of CSS features, due to implementation limitations;
for example, margins did not replace default margins, but added to them
(e.g., setting a top margin on a P element merely added to what we
already used as a top margin).

3) IE 3.0 had dual sets of rendering information - one for its previous,
"traditional" HTML rendering, and another tree-based set of rendering
bits that were only used for stylesheets.  This combination was (and I
freely admit this) a hack; but it allowed us to experiment with
stylesheet support without having to start from square one with our
normal HTML rendering.

IE 4.0 doesn't work like this - we DID start from square one with our
HTML rendering, and with a unified internal system (HTML attributes and
intrinsic renderings are treated almost identically to stylesheet
rules), so some of the side effects of the IE 3.0 hack are gone (e.g.,
inheritance works properly when you combine HTML "presentational tags"
and stylesheet rules), and some issues come up because of this - the
biggest one, obviously, is tables.  The choice we had was between
allowing inheritance of ANY rendering properties into tables (and
therefore, breaking the model we've had since the introduction of
tables, and breaking compatibility with millions of pages), and setting
up a set of internal rules that reset those rendering properties on
table cells.  Sorry for the inconvenience, but I had to vote for the
latter.  The only other solution would have been a hack that would
effectively separate the rendering properties by "set through HTML" and
"set through stylesheets" - and this is not, IMHO, a forward-thinking
strategy.

Again, my apologies - we have been investing in an architecture that
will be forward-compatible, instead of one that is tied down in legacy
rendering, while trying not to break that legacy rendering.  It's a fine
line, and wherever we tread, someone's going to be unhappy.

I won't dive into your comments on push technologies and Dynamic HTML -
whether intended or not, they're serious flame bait - except to
reiterate that we are, in fact, trying to get the fundamentals right.
They haven't been in the past - setting that aright without burning
backwards compatibility altogether is a difficult (and ongoing) task.

	-Chris
Chris Wilson
cwilso@microsoft.com
***

> -----Original Message-----
> From:	Derek Harding [SMTP:derek@tpd.com]
> Sent:	Saturday, October 04, 1997 6:55 PM
> To:	www-style@w3.org; Liam Quinn
> Subject:	Re: CSS1 and tables
> 
...
> Here I am trying to get my team to write good HTML and we're thwated
> at
> every turn by inconcistencies between different browser companies and
> even
> between browser versions from the same company. (Obviously the latter
> doesn't apply to Netscape coz they were too busy implementing their
> own
> "features" to get round to style sheets until version 4.0)
> 
> Do I sound cynical about NS & MS yet? Sorry, I'm just feeling that way
> atm.
> Huge amounts of effort is going into trying to beat their opponent
> (eg.
> different push technologies, different DHTML definitions etc.) it
> seems that
> they don't have time left to make the fundamentals work properly. :(
> 
> Derek
> 
> ---
> Derek Harding
> Technical Director, TPD Publishing
> http://www.tpd.com/~derek/
> ______________________________________________________________________
> ____
> "It's going to look pretty good, then, isn't it," said War testily,
> "the
> One Horseman and Three Pedestrians of the Apocralypse."
>         -- Terry Pratchett, Sourcery
>

Received on Sunday, 5 October 1997 16:58:57 UTC