W3C home > Mailing lists > Public > public-wai-ert@w3.org > May 2005

Re: more properties for the location class (was Re: locating assertions in non-text based formats)

From: Nick Kew <nick@webthing.com>
Date: Mon, 02 May 2005 00:52:19 +0100
Message-ID: <42756BB3.7040001@webthing.com>
To: public-wai-ert@w3.org

Shadi Abou-Zahra wrote:
> Hi Nick,
>>What's the current status of earl:location?
> Happy this caught your attention, I think you may have some valuable
> input to this discussion. Basically we've identified the need to be able
> to "locate" the occurence(s) of an earl:result within an earl:subject.

Indeed.  This is a long-standing issue I've written about before now.
A quick google finds the most important references:


Jim's fuzzy pointer document is a much briefer summary of some of the
things we were looking at around that time.

> Our first attempt was to look at XPointer/XPath expressions which work
> very well but only for some type of documents. Per specification, XML
> formats only but some parsers extend that to DOM as well. Jim also just
> posted info about fuzzy pointers but that too has document format
> limitations.


But one thing that bothers me badly about all these measures is that
none of them are robust against changes in a document.  Accessibility
audit requires us to flag up when a change to a document or template
does or doesn't materially affect the current accessibility report and
thus potentially call for reevaluation.

I'm wondering if it's worth resurrecting my old normalised
significant-change-detection script
http://lists.w3.org/Archives/Public/w3c-wai-er-ig/2002Jan/0019.html ?

> The big question remains, how do we locate the occurence of results in
> Web content that may be non-markup or even non-text based? Can we say
> more than "somewhere inside this subject there is an error"?

My tools, and I believe others, have hitherto used homegrown
descriptions of exact locations in a document.  Of course a
standardised vocabulary is preferable:-)

> At some point, we thought we may be better introducing all types of
> pointer possibilities (such as line numbers etc), as well as a
> processing model for tools of when/how to use these properties.

"all types of ..." rather begs the question.  Speaking as someone with
a background in pure maths, my instinct is to avoid any solution that'll
never be more than partial, by working around the question as somewhat
discussed in the first two references above.

> first discussions on this are at:
> <http://lists.w3.org/Archives/Public/public-wai-ert/2005Apr/0001.html>
> Note, this contribution may also be interesting backgroud:
> <http://lists.w3.org/Archives/Public/public-wai-ert/2005Mar/0026.html>

I wouldn't hold your breath for that to happen anytime soon.  But of
course, whatever work ER do in this area should feed back to QA-dev
and the validator team, and vice versa.

Nick Kew
Received on Sunday, 1 May 2005 23:51:59 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:52:49 UTC