Re: media characterization and formal types in HTML and CSS2

Another difficulty with Al's approach is that an HTML user agent which is
4.0 compliant, but which does not support any style language, would be
unable to take advantage of such expressions as the following (this
is an approximation of one of the examples given in the 4.0 draft):

<link media="print" href="document.ps" rel="alternate">

Note that if certain base types, such as "print", "screen" etc., are
defined within HTML, there is no need for style support, or for dependence
on any particular style language, in order for the above expression to be
processed correctly.

Al is also correct in acknowledging that, according to his strategy, a
media type unaware style language will preclude the user agent from
determining the applicability of the style sheet to the output medium
before fetching the style sheet itself. I suspect, though can not confirm,
that DSSSL does not include provision for media types as such, and thus
the disadvantage just outlined may produce inefficiency in practice more
often than might be initially apparent.

The solution that I would like to propose adopts a midway course between
exhaustively defining media types in the HTML specification, and
completely letting go of such responsibility.

The HTML specification should first present the rules for correctly
parsing the MEDIA attribute. These rules were briefly set out in the
initial draft but have been dropped in the specification itself. It should
then be stated that an HTML user agent should be able to recognise the
following media types, irrespective of the style language being used [list
of fundamental base types]. This provision is designed to overcome both of
the limitations that were discussed earlier in this message. The
specification would then go on to explain that a conforming HTML user
agent may recognise additional media types, including such new base types,
and extensions to existing base types, as are defined in the currently
active style language, or, if supported, the style language which is
designated by the internet MIME type of the style sheet to which the link
refers.

This exposition is admittedly somewhat inaccurate but I hope that it
suffices to communicate the general scheme.

In closing I would like to remark that Al and I are perhaps not too far
from agreement on this issue.

Received on Saturday, 20 September 1997 20:30:13 UTC