W3C home > Mailing lists > Public > www-html@w3.org > March 2003

Re: Against 'start' and 'value' attributes

From: Tantek Çelik <tantek@cs.stanford.edu>
Date: Mon, 10 Mar 2003 22:49:28 -0800
To: Daniel Glazman <glazman@netscape.com>, <www-html@w3.org>
Message-ID: <BA9255B9.223E9%tantek@cs.stanford.edu>

On 3/10/03 7:29 AM, "Daniel Glazman" <glazman@netscape.com> wrote:

> Etan Wexler wrote:
>>> If I receive a text with questions numbered 1 to 6 in a list
>>> and only have answers to 2 to 4, I may want to start a list with start="2".
>> Sure, I agree that one may want to start a list at a number other than 1.
>> What I don't agree is that a 'start' attribute is a good solution. What if
>> you have only answers to 2, 4, and 5? With a simplistic 'start' attribute,
>> you cannot express the correct information.
> Strange answer, indeed. In summary, you tell me that my argument is not
> relevant because users may as well want to type lists than do not
> represent a range. I agree; but that does not suppress the fact they
> also may want to do ranges!

Yes Daniel, your analysis is correct.

Unfortunately this flawed argument style that you have dissected appears to
be used a lot in W3C lists.

>> An element of a list-marker type
>> accomodates both this case and many more with facility. Consider:
>> <ol>
>>  <li><marker>2.</marker> ...</li>
>>  <li><marker>4.</marker> ...</li>
>>  <li><marker>5.</marker> ...</li>
>> </ol>
> In an ideal world, we would have IDREFs and CSS styles allowing to do
> everything. We're not in that world.


>>> Not all instances of lists with a start index > 1 are continuing lists.
>> I'm glad that you raised this point. I had failed to consider it. But now,
>> having considered it, I remain at my original position. I do not think that
>> a 'start' attribute is the solution, as noted above.
> Still disagreeing.

And I agree 100% with Daniel on this.

>>> The "let's do that with CSS" answer is not more satisfying. The fact
>>> that the list's numbering starts at 2 instead of 1 is not purely
>>> presentational but also content-related.
>> I half agree. Numbering becomes an essential part of content in cases where
>> hypertext does not apply.

This is false.  Numbering and hypertext _can_ be codependent, but they can
also be completely orthogonal.

>> Printed documents like legal briefs are prime
>> examples. And, without a doubt, people will use XHTML 2 to comment on
>> print-only documents. Numbering therefore remains a real concern.
> Woof. From 1991 until now, the number of "without a doubt" predictions
> that finally became true is imho very limited...


>> What I must emphasize, though, is that reference by number (or other mark)
>> is a leftover of non-hyper text. In a pure hypertext, reference would be by
>> links. If, in your example, links connected each answer to the corresponding
>> question, a user agent could follow the links and present the corresponding
>> list item markers with your answers.
> Right. IDREFs. See below.
>> In pure hypertext, numbering typically is presentational. The
>> machine-readable link forms the stable basis of the reference mechanism
>> available to the user. The particular numbers used as labels become
>> unimportant; even if the numbers change in the original, the change
>> propagates effortlessly through all affected documents.
> That's far too idealistic. Imagine a list of questions but they don't
> have IDs, you can't target them. You can't modify that because that part
> of the document is readonly. What do you do ?
> </Daniel>

I agree with Daniel's sentiment and his original message.

There are absolutely uses for start and value attributes.  I can see marking
up the markers with tags rather than attributes, but certainly in the
immediate future it makes sense to undeprecate start and value while a
better solution is worked out.  I see no problem with having both solutions

Received on Tuesday, 11 March 2003 01:36:29 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:06:03 UTC