[conformance techniques] How to connect to an EARL claim.

Returning to "where does the [reference to the] EARL go?" topic.


http://lists.w3.org/Archives/Public/w3c-wai-er-ig/2001Oct/thread.html#23

1. Suppose a page incorporates a brief compliance statement (e.g. 'valid'
logo).

a. If the logo is in smart media, PNG or SVG, then a link [leading down a
link path] to the EARL equivalent can be embedded in it.

b. If the logo is in dumb media, then the path to the logo can be
programmed through a semantic wrapper which contains both a 'load
immediate' reference to the dumb media image and passive links to more
information including the EARL.

c. The path through the wrapper can be included as a 'try harder' option in
OBJECT if we can negotiate appropriate OBJECT semantics with the HTML WG
such that the ordered search-list method of resource alternative selection
is recognized as an optimized hint from the author and not the only way
that User Agents (or servers) may select.  It should also be legal (in
order to be consistent with UAAG Guideline 2) to select by client-side
application of preferences-based rules or by interactive choice.  This
allows the user to benefit from the searchlist order preferences as known
by the author without having any resources rendered inaccessible by the
unwavering mechanical application of the author's understanding of the
preferability of different alternatives, which the author cannot know
correctly for all users.

d. The path to site or page conformance claims can flow from a "usageNotes"
or "accessibilityClaims" or "more about me" LINK to a multi-lingual sheaf
of documents multiplexed by RDDL.  This can be to information about the
page or the site or anything else containing the page which is a good
notional anchor for the meta-information.  Within the EARL view the subject
of any claim will be unambiguous.  The EARL is about the current page or
some scope containing the current page and in which the connection with the
current page is not that hard to trace.

2. The page contains no "show by default" accessibility or other
metadata-reference indication.

All of the above apply, except that the metadata and assertions associated
with a site or other pageAffinityGroup logo will likely address more than
just accessibility.

What is required in terms of rel/role vocabulary is that there be a simple
'meta' or 'pertainsToMe' semantic defined and that more precise
relationships be formally connected to this via suitable application of
is_a and _describes_ relationships.  See Larry Masinter's "that described
by" URI variant for something which encapsulates the two assertions.

For example A describes properties of a category X.  B is_a X.  This can be
articulated without the indirection through the formal variable by
concatenating the relations:

B is_a [ arcDescribes [A] ].  Or A _describes_ B.  But if B is to be coined
after A, then A must couch its assertions not on instance identities, but
on accept-patterns the implicitly define a class that A may conform to.

We don't want the semantic of the reference off the normal page to be
'conforms to' which requires a black-and-white condition.  It is useful to
link to data that casts information on the accessibility, and not just
black and white conformance claims.  A clear level here is reading level
measurements.  Metadata giving measured reading-level grades for content is
very useful even 'though we don't have a universal scale for measuring it
or a Web-wide standard for what it should be.  The assertion of conformance
to boolean conditions is included in the function "gives information about
[topic]" and the distinction as to what mode of information giving [boolean
or statistically correlated] should be distinguished in the cited "more
information" utterance and not in the language semantics of the citation.
This may be hard for people to get used to, but it is the right way to do
things.  We can have more precise relationships, but they have to be is_a
threaded to the loose one and the architectural framework works off the
weakest.

Al

Received on Sunday, 18 November 2001 14:46:48 UTC