Re: Mandatory and Important

At 12:44  -0400 22/08/08, David Poehlman wrote:
>most sane tools should warn that it is missing and not pass it forcing the
>proper attention to it.  @role still does not cover what can only be human
>judgements till we get machines that can parse images and make textual sense
>out of them.

You cannot assume that the person providing input to the tool that 
makes HTML was the author of the input, and you cannot assume that 
there is anyone to report to at the time the HTML is being made.

Consider a hypothetical service which takes word documents in a drop 
box and later makes HTML pages.  The generation happens after the 
drop, the person dropping the document may not have written them 
anyway, and word has no field for alt text of images.  The tool 
should make functional, comformant HTML.  Clearly it cannot invent 
alt text, and so it cannot conform to WCAG, but we should at least 
make sure that consumers of the HTML are not *mis*-led although if 
they cannot see images there may be semantics missing.

alt="" role="unknown"
alt="{auto-converted}"

something else?

>
>----- Original Message -----
>From: "Dave Singer" <singer@apple.com>
>To: "David Poehlman" <david.poehlman@handsontechnologeyes.com>; "Al Gilman"
><alfred.s.gilman@ieee.org>; "HTML WG" <public-html@w3.org>
>Cc: "Doug Schepers" <schepers@w3.org>; "Karl Dubost" <karl@w3.org>; "W3C
>WAI-XTECH" <wai-xtech@w3.org>
>Sent: Friday, August 22, 2008 12:30 PM
>Subject: Re: Mandatory and Important
>
>
>
>At 10:05  -0400 22/08/08, David Poehlman wrote:
>>not optional, missing.  If it is missing it breaks spec but is still
>>missing
>>so tools/authors need to fix it so that it is not missing.  The {} for
>>instance was the hack to prevent missing.  I am saying that it should be
>>real.
>
>But if missing is non-conformant, most sane tool authors will insert
>it to avoid a conformance failure.
>
>Then they insert alt="" (a lie) or alt="random text" (useless).
>
>Look, honestly, I don't want to sound harsh, and I value the
>dialogue, but until someone is actually willing to provide an
>alternative answer to the question -- not duck it, change it, or deny
>the problem exists -- we are just annoying each other.  The spec. at
>least contains *an* answer, and it seems as if the discussion of role
>might converge on another.
>--
>David Singer
>Apple/QuickTime


-- 
David Singer
Apple/QuickTime

Received on Friday, 22 August 2008 16:58:22 UTC