Because, as a potential consumer of the product, it "feels wrong" to me.
Stylesheets feel right, but the inability to put information locally,
when appropriate, feels wrong. Intuition, based on experience, tells
me I will want to put my fingers where your restriction says I can't...
That's the trouble. You are attempting language design based on your
speculations about perceived needs. As I have said very early in these
threads, one shouldn't design compromise into a language (or anything
else) unless one had either good empirical data on hand or clear historical
compability requirements. I don't believe that historical requirements
can be invoked here so we're left with empirical requirements for which
we have no real data. Given this fact (and I'd challenge anyone to
effectively dispute this as a fact) and given that the STYLE attribute
as has been discussed violates basic principles on at least two counts:
(1) don't provide multiple ways of saying the same thing when one will
suffice; and (2) don't mix form and content, I'd suggest that the authors
of DRAFT-IETF-HTML-STYLE-00.TXT remove or modify their proposed use of
a STYLE attribute as a binding mechanism. A modification which would
be acceptable to me is to specify the declared value of the STYLE
attribute to be either NAME or NAMES.
Regards,
Glenn Adams