W3C home > Mailing lists > Public > w3c-wai-ig@w3.org > January to March 2014

Re: H43 and header cell relationships

From: Olaf Drümmer <olaf@druemmer.com>
Date: Thu, 20 Feb 2014 15:32:58 +0100
Cc: Olaf Drümmer <olaf@druemmer.com>, W3C WAI ig <w3c-wai-ig@w3.org>, Matt Tongue <Matt.Tongue@nrc-cnrc.gc.ca>
Message-Id: <F7173847-12E1-4B9E-8020-C209F893858C@druemmer.com>
To: Jonathan Avila <jon.avila@ssbbartgroup.com>
Hi Jonathan,

PDF, in its definition for Tagged PDF, had it right since 2001:

about the Headers attribute in ISO 32000-1, Table 349 – Standard table attributes, page 610:
> This attribute may apply to header cells (TH) as well as data cells (TD) (see Table 337). Therefore, the headers associated with any cell shall be those in its Headers array plus those in the Headers array of any TH cells in that array, and so on recursively.
[this is also the reason I thought that HTML just does it the same way, but seemingly I was wrong (or maybe not - maybe HTML4/5 does not explicitly prohibit nested use of the headers attribute?)]

Having done a lot of research about tables in the last few months, I am by now highly confident that PDF conceptually does the right thing, and that what HTML suggests is conceptually inappropriate (and hinders reliable use of the headers attribute). Once I find the time I will put together a paper illustrating this but currently I am too busy with my day to day job….


On 20 Feb 2014, at 14:51, Jonathan Avila <jon.avila@ssbbartgroup.com> wrote:

> Olaf, the first step I would do is to try to validate or disprove the
> possibility of a nesting/serial access approach.  Is it possible to design
> a table where a references b and b references c but a should not reference
> c?  If that's possible then you can't rely on this serial type approach.
> Jonathan
> -----Original Message-----
> From: Olaf Drümmer [mailto:olaf@druemmer.com]
> Sent: Wednesday, February 19, 2014 5:19 PM
> To: W3C WAI ig; Matt Tongue
> Cc: Olaf Drümmer
> Subject: Re: H43 and header cell relationships
> I think I have to correct my statement:
> HTML seems to require (as Sailesh points out) that indeed the headers
> attribute of table cell has to list the ids of all the header cells with
> which it is associated.
> Sorry for any confusion my statement might have caused.
> I nevertheless would like to add that I think that this approach is
> conceptually wrong - nested semantic structure should be expressed via
> nested representation of pertinent data / attributes. Even if one were to
> agree though with my reasoning it's probably a bit late in the game to
> have this addressed...
> Olaf
> On 19 Feb 2014, at 22:46, Olaf Drümmer <olaf@druemmer.com> wrote:
>> As far as I can tell, any cell should identify only its direct header
> cell parents. Nested header cell relationships would then be represented
> by nested use of the headers attribute.
>> Olaf
>> On 19 Feb 2014, at 22:09, "Tongue, Matt" <Matt.Tongue@nrc-cnrc.gc.ca>
> wrote:
>>> When marking up a table with multiple levels of headings, is it
> mandatory to always put all header cell IDs into the headers attribute of
> a cell?
>>> For example, if a data cell has 3 header cells, but the 2nd header
> cell's headers attribute references the 1st header's ID, would it not
> suffice for the data cell to just reference header cells 2 and 3, since
> header cell 2 references header cell 1 (thus creating a relationship
> already)? Or must every header always be identified for every cell, no
> matter what?
>>> Reference: http://www.w3.org/TR/2013/NOTE-WCAG20-TECHS-20130905/H43
Received on Thursday, 20 February 2014 14:33:26 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 13 October 2015 16:21:50 UTC