Re: Thoughts on CLASS and ROLE attributes in HTML 3 draft

Peter Flynn (
11 Mar 1995 01:11:00 +0000 (GMT)

Joe writes:
I think it would be a good idea to orthogonalize these attributes;
that is, leave CLASS attribute values entirely in the user's
namespace, and use ROLE for everything with predefined semantics.

How does that sound?

Pretty good. I've just been thru the DTD in more detail...Dave's done
his usual excellent job...coupla queries:

a. Can we insert the RFC # for the ref to HTTP under HTTP-Method?

b. <!-- The style attribute is used to subclass HTML elements for

c. lang CDATA "" -- ISO language, country code --
What ISO # defines these? 3166 is country codes, what is langs?
Does anyone have a full list they could post or point at?

d. <!-- SGML standard forces different NAMES for all attribute values
Could someone run this by me again?

e. 'clear CDATA "no" -- (left|right|all|no) move down past figures --
"Move down past figures" doesn't explain exactly what is intended
here. Can you give a couple of examples?

f. <!-- ENTITY % needs
Ought to have "units" rather than trying to embed them in "needs"

g. I'm not familiar with MD5. To me, a "message digest" is something
like a digested mailing list. Is this intended to hold an
arbitrarily long string of base64 which the browser will decode and
display? Where would a user get some text encoded into base64
without pushing it thru a MIME mailer and snipping it out?

h. HTMLat1
If this is going to include all of ISOLat1, then I think we should
reference ISOLat1 separately, and keep HTMLat1 (named differently)
for our subset of ISOpub and ISOtech.

i. mdash
typo in expansion says [ndash ]

j. the widely available Adobe math font
What font? What typeface? Remind me to spew out the LaTeX list.

k. The ALIGN attribute is used for math.
How do you specify inline vs. display math in HTML3? Or do you have
to put it inside a <p align="center"> ?

l. Does nowrap in <p> effectively imply auto-<br>?

m. Mixed content problem in %block; sayd: is that it also prevents:
"<LI> <P>some text" since it forbids PCDATA What I can't see is why
anyone would want to say <LI> <P> with a space between them in the
first place: if you want a para inside the list item, then just
stick the <p> right after the <li> with no space. While it's handy
sometimes to be able to have <p>s inside <li>s, I think there are
many more occasions when people will want text content directly
inside <li>

n. DL
I still think HTML.Recommended should be (((DT,DL+)+)|((DT|DD)+))+
eg you can either mix and match as you please, or be stricter and
have each DT followed by 0+ DDs. Is that right?

o. should we allow a list header?
Yes, definitely.

p. e.g. dingbat=folder
Doesn't this have to be dingbat="&folder;"

q. Is it possible to say <!ELEMENT DIV - - (%body.content;|html)> ?

r. Where is this RANGE control implemented? Is SIZE="1,10" meant to
imply limits:min=1,max=10 ?

s. I didn't find FLOAT used, either.

t. CAPTION alignment: top/bottom should be separate controls from
left/right. What happens if you want the caption in the TLHC?

u. If we have <tr class=...> do we still need TH and TD?

v. Figures: you can request the browser to move down until there is
enough room...I'm being very dense here: move _what_ down? the
figure? or the preceding element? or the following element? I find
this use of "move down" very confusing. If we want text to flow
round a figure, we should say <figure textflow="left"> or something
explicit. The use of "clear" also carries the wrong semantic
connotation IMHO...can we not find a better word?

w. BASE has presumably been readmitted to the fold?

x. Do we have real-life examples of META yet?

y. Statements in the given style notation
Have we defined whata style sheet looks like yet? Or are we
sticking to DSSSL-lite? I've tried to print the DSSSL .ps doc, but
it gags my printer. Is there a quick-start guide to DSSL?

I've just used Near&Far to do a diag of the HTML3 structure which I'll
upload over the weekend, also a sheaf of its reports on element,
attribute and entity usage, including some excellent x-ref
listings. If we could make sure every element, attribute and entity
has a good explanatory comment after it --like this-- then that will
get picked up by N&F and we can use it for these tables. I'm going
away for 3 days Sunday, so if I don't get to do this upload in time,
someone nag me next week.