RE: [techs] The TH Rule

Okay, so I've been mostly lurking...

Gregg asked:
> Any specific recommendations for out guidelines?

I think Chris' good work provides justification for making things much simpler.

1)  Data tables always use TH.
2)  Layout tables never use TH.

This makes the situation easy for user agents.  The difficulty is for human authors who might not be sure themselves if they are writing a data table or layout table.  Generally, we give authors the benefit of the doubt when deciding what is "good" ALT content.  For some tables the decision that it's layout or data will be a judgment call.  Most of these situations will be relatively straight forward (just two or three columns or rows) so hopefully they are not too problematic irregardless of the presence of TH.  For the more confounding examples Chris cited, might the recommendation for nesting tables be a solution?  Skilled use of HEADERS and ID should also work, so is there harm in declaring that mixed tables are data tables?

We still need to work on the heuristic for determining when a table is for data table and when it is for layout.  IMHO it is a little too strict to just say that only when cell contents can be read out of order is it a layout table.

Carlos wrote:
> In my view, an empty "alt" attribute is a value. As you know, in many 
> programming languages, a "null" string is not the same as an "empty" string.

Of course not.  That is not the relevant portion of the spec that Chris quoted.  The spec says:
>> you need to provide a description with ALT

We routinely try to enlighten people that description is NOT usually good content for ALT.  But there is that unfortunate advise in the spec.

Reliable TH use/avoidance would make things so much easier for user agents that it definitely belongs in WCAG2 IMHO.  The other issue, of course, is that automated checkers don't dare rely on the presence (or lack) of TH to distinguish between data and layout tables.

Jim suggests:
> Always use summary or caption on data tables.
> Don't use summary or caption on layout tables (please).

But that would make all those charming 
summary="table used for layout only"
code fragments we see just plain wrong...

Seriously though, currently many WCAG1-A pages might well use layout tables with summaries.  If they were purposely designed to be accessible, they probably appropriately use/avoid TH.  We want WCAG2 to be backwards compatible, so recommending against summary (for layout tables) compromises that.  The presence (or lack) of summary might help distinguish between data an layout tables, but it doesn't help _navigate_ the table.  Summary helps the user but not the user agent.  TH really is a tool for the user agent.

Received on Tuesday, 11 May 2004 13:49:20 UTC