W3C home > Mailing lists > Public > www-html@w3.org > July 2000

Re: Alt attribute for unimportant images

From: Jan Roland Eriksson <jrexon@newsguy.com>
Date: Thu, 13 Jul 2000 20:08:51 +0200
To: Dave J Woolley <DJW@bts.co.uk>
Cc: www-html@w3.org
Message-ID: <s41sms89o1dmu2eq371bm0g2j5pkt0rkd9@4ax.com>
On Wed, 12 Jul 2000 18:35:04 +0100, Dave  J Woolley <DJW@bts.co.uk>

>> From:	Jan Roland Eriksson [SMTP:jrexon@newsguy.com]
>> Which is using a <!DOCTYPE... declaration for a purpose it was never
>> designed to handle. That use of <!DOCTYPE... is wrong and I have made
>> that view of mine very clear on several occasions.

>	[DJW:]  
>	As the DTD cannot provide any semantics information
>	other than by the user agent matching the public (or
>	system identifier) againts a set of semantics rules,

I did not know that you could "walk a slack rope", be careful not to
fall because I can't see a safety net :-)

Still it would be interesting to find out where in UA's you can find
"sets of semantics rules". Todays (and tomorrows) "popular browsers"
only displays stuff on a VDU for Gods sake. Well some of them can print
on paper or run projections too, for better or for worse, but that does
not change the basics.

There is no semantic definitions what so ever, in a traditional SGML
type DTD, and styled presentation in a UA can not add info about the
meaning of a documents content either.

The only SGML way to convey the _meaning_ of marked up content is to
properly define an _application_ of SGML, where an SGML declaration and
a DTD are just rather small parts of the full size application.

The lion share of the work lies in writing the formal prose that
describes the characteristics of each defined elements allowed content.
That's where semantic definitions comes in, nowhere else.

E.g. the "Text Encoding Initiative" (TEI) and DocBook are prime examples
of properly defined applications of SGML. And I will stretch as far as
to say that RFC1866 falls into the same category too, it was a good
attempt to define HTML2 as a true application of SGML.

>	it seems to me that it is a perfectly reasonable use of
>	<!DOCTYPE to tell the user agent what semantics to apply
>	to the data (or confirm that it is an unknown type).

Don't confuse "presentation" with "semantics" please.

If any one at all knows what a www document contains, and what would be
the best suggestion on how to render it, in a UA with multiple rendering
modes available, it's the document author.

So, give me as the author, a possibility to send an unambiguous
suggestion to the UA for what rendering mode to use.

That can easily be done through an HTTP extension header that the UA
follows at priority level 2, after the users "hard and definite"
selection at priority level 1, if s/he wants to make one.

At priority level 3 we could add handling of a <META HTTP-EQUIV... to
the same effect, for those authors that are so unlucky that they can not
configure their servers themselves.

At priority level 4, use <!DOCTYPE... sniffing then if it has to be
there, but personally I see priorities 1-3 as fully sufficient.

Jan Roland Eriksson <jrexon@newsguy.com>
Received on Thursday, 13 July 2000 14:08:15 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:05:54 UTC