W3C home > Mailing lists > Public > public-html@w3.org > June 2009

Re: Summary of Thursday's IRC conversation about @summary

From: Leif Halvard Silli <lhs@malform.no>
Date: Tue, 09 Jun 2009 16:13:40 +0200
Message-ID: <4A2E6E14.80608@malform.no>
To: Henri Sivonen <hsivonen@iki.fi>
CC: Shelley Powers <shelleyp@burningbird.net>, "public-html@w3.org WG" <public-html@w3.org>
Henri Sivonen On 09-06-09 13.46:
> On Jun 9, 2009, at 14:31, Shelley Powers wrote:
>> Henri Sivonen wrote:
>>> I'm observing that in the light of 'best practice' how @summary is 
>>> supposed to be used, only the empty summary attributes comply but 
>>> actual longer-than-caption summaries of data tables are nowhere to be 
>>> found in Philip's data. (summary="Calendar" may be useful but doesn't 
>>> match the best practice put forward.)
>>
>> I don't believe a requirement for summary is that it must be longer 
>> than caption. Your argument is looking for a fault where there is none.
> 
> Quoting from http://www.w3.org/mid/4A279FC9.4080904@cfit.ie
> "It is also worth noting that <caption> is a terse descriptor. @summary 
> is a long descriptor."

Henri pointed to my message ...

> See also http://www.w3.org/mid/4A27C273.9030704@malform.no

... were I list everything said about @summary in HTML 4.

Reading those HTML 4 excerpts, it seems unfair to say that HTML 4 
_defines_ it as a long descriptor. Firstly, it does not say "long" 
but "longer description". Only in the "Changes" appendix does it 
use "long".

The DTD definition found in HTML 4 is "purpose/structure for 
speech output". (The "purpose" part should be dropped, see below.)

There is one wording in HTML 4 that goes like this:

     "A longer description may also be provided (via the summary 
attribute) for the benefit of people using speech or Braille-based 
user agents."

The important part of that sentence is "for the benefit". Evidence 
seems to suggest that, in general, the way to get benefit is to 
have the @summary describe the table _structure_, and nothing 
else. @summary should not eat into the purpose of <caption>. It 
should also not serve the purpose of a longer description of the 
table.  (Longer descriptions should be kept outside the table, and 
eventually be linked to via aria-described etc.)

Thus, it seems like HTML 4's DTD definition of @summary should be 
corrected/become more narrow: the "purpose" part should be dropped 
and only the "structure" part should be kept. What is meant by 
"structure"  should be defined and illustrated in the draft, so 
that authors can get a clear understanding of what is meant.

Thus, any use of @summary for captioning purposes shall have no 
basis in HTML 5. (This is in contrast to how @summary is often 
used. It is also in contrast to WCAG 2.0, which in Technique H73 
advices that @summary may also include captioning content, when or 
if the table is lacking a <caption>[1].) HTML 5 should also not 
allow for "creative" and "over-helpful" summaries - it has been 
relayed to me by an expert that such application of @summary is 
only confusing. HTML 5 should only allow for neutral table 
structure descriptions. Such structure descriptions will also 
"last longer" (if the table gets updated) and thus less easily get 
out of sync with the table content.

A particular usecase: One of the examples discussed in this group 
some months ago [sorry no link] was a web page that used the 
@summary to convey the sort order of a table (the sort order could 
be changed interactively - it was a search result table). That 
seems like an good example of what @summary could be used for. 
Admittedly, such info could perhaps in theory also be added to the 
<caption>. However, it seem illogical to change the caption, 
because a change of the caption would indicate that "this is 
another a table than the previous you saw" rather than "this is a 
new state of the same table". To sighted users it is only 
confusing if some text changes for no obvious reason, as it means 
that they have to reread the text and create a "diff", in order to 
see if the text says something material.

I speak only on behalf of myself. But I think the above reflects 
an emerging common understanding within the HTML4all.org 
community. (Especially many thanks to Gez Lemon for a clear vision!)

[1] http://www.w3.org/TR/WCAG20-TECHS/H73
-- 
leif halvard silli
Received on Tuesday, 9 June 2009 14:14:21 UTC

This archive was generated by hypermail 2.3.1 : Monday, 29 September 2014 09:39:04 UTC