- From: James Graham <jg307@cam.ac.uk>
- Date: Wed, 13 Jun 2007 16:15:41 +0100
- To: "Gregory J. Rosmaita" <oedipus@hicom.net>
- CC: public-html@w3.org
Gregory J. Rosmaita wrote: > james wrote, quote > I don't actually know the reasons for not including summary but I would > guess they were related to the fact that @summary is tying up important > metadata in an invisible attribute. Invisible data is more likely to be > missing or incorrect than visible data; according to [1] @summary is > present on about 2.5% of tables, I would expect it to be unhelpful on > many of those (though clearly some actual research would be needed here. > [2] has some results, mostly from CMS templates). > unquote > > GJR: i think the search parameters you used are too limited - summary > may or may not be the first attribute for a TABLE, but it usually > isn't, so searching for summary=" yields more results Actually the code search seems to return cases where summary is not the first attribute as well. In any case, I was more concerned with the quality of the data in @summary than the number of results. Since I use a graphical UA I don't have a feel for what/how much information is needed for a summary to be useful, so I'm not the best person to evaluate these results (I also think the survey is highly biased). > james also wrote, quote > OK, this sounds like a use case for a feature providing an overview of a > table's contents. In general I think it's better to work from use cases > + backward compatibility requirements rather than HTML4 directly. It > seems like some of the problems could be solved automatically e.g. > saying how many rows and columns are in the table. Is this not the case? > If this can be done, the remaining problem sounds like it could be > solved either through <caption> or another mechanism for associating > text that is, by default, displayed in visual UAs with the table. Of > course I also believe that the behavior of @summary should be specced > for UAs. > unquote > > GJR: it's not just a question of raw statistics - my screen-reader tells > me that it is X columns by Y rows, but that doesn't alert me to the fact > that there may be two rows of table headings the first spanning all the > TABLE's columns, the other only a select few (tables such as the *** > example on andre's test page [note *]; the more nested tables, the worse > for the screen-reader user, as one has to traverse and inspect every > table between the first table and that nested table for which the user > is searching Is it conceviable that this process could be automated by the screenreader scanning the <th> elements and using a combination of their position, their scope attribute (if present) and their row/col spans to deduce the number of columns and rows? > as far as the wiki page HTML/SummaryForTABLE, do you want your comments > and suggestions slash questions under reasons for deprecation, under > work-arounds, or raw data? in any event, i will add your post to the > list of posts on the topic... Well I think the fact that @summary is explicity invisible metadata is a reason to remove it, so that argument should go into that section. I think explicit linking of (default) visible text content to a table should be suggested as a possible alternative to the summary attribute. -- "Eternity's a terrible thought. I mean, where's it all going to end?" -- Tom Stoppard, Rosencrantz and Guildenstern are Dead
Received on Wednesday, 13 June 2007 15:19:03 UTC