Re: Draft change proposal on location of alt guidance

On Thu, Jul 14, 2011 at 3:12 PM, Michael Cooper <cooper@w3.org> wrote:
> the requirements and use cases driving the history
> up to now remain unmet

These should be specified rather than alluded to, as the purpose of
the "Rationale" section is to state explicitly "What problems does the
proposal address".

http://dev.w3.org/html5/decision-policy/decision-policy.html#escalation-step-2b

> while new problems have been introduced:
>
> The two resources overlap and potentially contradict.
> The advice is overly prescriptive for a content language specification.
> The advice constitutes "best practices" guidelines that is applicable to a
> wider range of content technologies than HTML.
>
> Problem 1 seems self-evident, and the remainder of this rationale focuses on
> problems 2 and 3.
>
> W3C produces a variety of materials that play different roles. Two roles
> that are often confounded are content language definitions, and authoring
> best practice recommendations or guidelines. A content language
> specification defines the features of a content language, its lexical form,
> and processing expectations, while authoring guidelines suggest how best to
> take advantage of the content language features to achieve particular goals.
> There is a tendency to develop the two together, since content language
> features need to meet the use cases that drive authoring guidelines.
> Structually, however, the two types of document should be kept separate.
> When one specification includes both kinds of information, it becomes
> difficult to tell what statements are in which category and introduces
> confusion into the normative model.

I'm not sure I understand this section.

You're claiming that mixing informative material for content producers
with normative and informative material for content producers and
content consumers makes it hard to tell which material is informative
material for content producers?

Can you give a concrete example of the sort of confusion that might
realistically arise? Can you give any evidence that such confusion
would be likely?

Why is such confusion not also likely for informative material for
content consumers or informative versus normative material more
generally? It seems common for W3C specifications to mix informative
and normative material.

The W3C's Recommendation on how to write specifications recommends
clearly distinguishing informative and normative content, and clearly
stating the class of products to which requirements apply:

http://www.w3.org/TR/qaframe-spec/#norm-informative-gp

http://www.w3.org/TR/qaframe-spec/#implement-principle

It doesn't say anything about using multiple specifications (still
less different working groups!) to do this.

> Small authoring suggestions are useful
> to help explain the intent or use of a content language feature, but these
> need to be clearly non-normative, offered as guidance only, and not allowed
> to grow into a full set of best practices within the content language
> specification itself.

Why does shortness make confusion about the normativity and target of
material less likely?

> In general, the HTML 5 specification observes this principle and focuses on
> defining the features of the HTML 5 language without overly prescribing
> authoring or user agent expectations. But the section on text alternatives
> is a notable exception. The confusion it introduces on what constitutes a
> normative part of the HTML 5 conformance model is a part of the problem that
> has led to difficulty agreeing on the appropriate content of that section.

Section 4.8.1.1:

http://dev.w3.org/html5/spec/embedded-content-1.html#alt

consists of normative requirements for the content of the "alt"
attribute plus some examples illustrating those requirements:

e.g. "When an a element that creates a hyperlink, or a button element,
has no textual content but contains one or more images, the alt
attributes must contain text that together convey the purpose of the
link or button."

So what's the relevance of the argument that including informative
guidance for authors is confusing?

Are we arguing that we should remove all document conformance
requirements from HTML5? Given the above, why not? (Note this *has*
been seriously suggested and considered in the past, this isn't meant
as a /reductio ad absurdum/.)

> Beyond the issue of confounding technical features with authoring practices,
> in this particular case the authoring practices are applicable to more
> content languages than HTML. The guidance for optimal text alternatives in
> the various contexts for images applies to any content language that
> provides images with support for text alternatives.

Better give examples of how the very same recommendations apply to
specific non-HTML languages. Some of them seem specific to HTML to me,
for example the conformance requirements for @alt for an image made up
of a group of other <img> elements:

http://dev.w3.org/html5/spec/embedded-content-1.html#a-group-of-images-that-form-a-single-larger-picture-with-no-links

> If each content language were to define its own authoring practices, there
> would be both repetition of substantially similar information, and high risk
> that recommendations for similar situations would be different across
> different content languages.

This seems to be a problem introduced by the need to snapshot
documents as W3C recommendations, since Living Standards can be kept
in sync by editors without requiring authors to either find
implementation guidance applicable to their specific language amidsts
a morass of implementation details for other languages, or apply very
generalized rules to their particular language.

Inevitably, snapshoting documents as W3C recommendations means they
will get out of sync with best practices (WCAG 1.0 is an illustration
of this).

> This would introduce confusion to authors and likely reduce the quality of
> text alternatives overall across the Web. For this reason, it is preferrable
> to provide the resource in a location where it can be referenced by all
> content languages for which it is applicable. Because the section referenced
> above relates to accessibility authoring practices for text alternatives, a
> resource produced by the Web Content Accessibility Guidelines seems to be
> the most appropriate location. This group has produced the W3C's
> technology-neutral accessibility authoring advice and maintains support
> materials including general techniques, technology-specific techniques, and
> interpretive guidance, so is well suited to maintain this document.

Having two different groups working on language features versus
accessibility guidance could make it more likely that language
features will be poorly designed
from an accessibility point of view. Maybe one for negative effects?

> Details
>
> Remove from the HTML specification the section 4.8.1.1 Requirements for
> providing text to act as an alternative for images.
> In the HTML specification 4.8.1 The img element, change the sentence "The
> requirements on the alt attribute's value are described in the next
> section." to "Recommendations for optimal values of the alt attribute are
> available in external resources, such as [Techniques for providing useful
> text alternatives]".

You're linking to a Working Draft at a URL that changes. In the final
form of the Change Proposal I recommend linking either to a permalink
or stating a revision of the spec as a baseline for changes.

> Note: The text in brackets is intended to be a link, URL to be provided
> later, which depends on how the full implemention of this change proposal is
> done. This link would be an informative reference, not a normative
> dependency.
> Note: The specific wording suggestion above is not as important as the gist
> of the wording, which is to inform readers where outside the HTML
> specification to look for guidelines on text alternatives, as part of the
> central goal of moving that information to external resources. Other change
> proposals may involve tweaks to the recommended wording that are not
> necessarily in conflict with the suggestion of this change proposal.

What are you proposing happens to the preceding sentence?

"Except where otherwise specified, the alt attribute must be specified
and its value must not be empty; the value must be an appropriate
replacement for the image"

"specified" refers to normative requirements.

I'm really unclear on what's being proposed here. Is it:

   1. To remove *all* normative authoring requirements around @alt
from HTML5? e.g. <img src="photo.jpg"> would be conforming HTML5.

   2. To require @alt, but remove all normative authoring requirements
around its contents from HTML5? e.g. <img src="pixel.gif"
alt="spacer"> would be conforming HTML5 but <img src="photo.jpg">
would not and <img src="photo.jpg" aria-label="Black cat Sophie
chasing her tail"> would not. This seems to conflict with another
change proposal being considered by the Task Force (making @aria-label
conforming in the absence of @alt)

   3. To provide normative authoring requirements for the presence and
contents of the @alt attribute in another document to be produced by a
different working group, such that authors, conformance tool
developers, authoring tool developers, and so on would be required to
consult that other document to determine if markup is conforming?

--
Benjamin Hawkes-Lewis

Received on Thursday, 14 July 2011 17:13:47 UTC