"Equivalency" in UAAG/WCAG, Checkpoint 2.3

To: UA List
From: Eric Hansen

In this memo I hope to address issues regarding "'Equivalency' in UAAG/WCAG, 
Checkpoint 2.3" [1] and succeeding memos by Al Gilman ([2], [4]) and Ian 
Jacobs [3]. I hope that this will clarify things and move toward a 
resolution.

INTRODUCTION

Ian's discussion [3] and Al's response [4] have provided a good discussion 
of senses in which equivalents and equivalency targets should convey the 
same meaning. Both seem agreed that the type of equivalence found between an 
equivalent and an equivalency target is usually approximate and that Web 
content authors should work toward having the equivalence as exact as 
possible. Al's discussion of "equality" versus "equivalence" seems to 
correctly capture some of the differences in connotation between the two 
terms. I can see how an "equivalency relationship" might be thought of as 
perhaps a less exact form of "equality relationship".

As I have thought about our present discussion, I think that there is an 
opportunity to reframe the discussion in a way that shows a pathway out of 
our disagreements. Specifically, I would like to do so my focusing more 
precisely on an important area of symmetry to supplement our ongoing 
acknowledgments of asymmetry.

A Critical Symmetry Between the Equivalent and the Equivalency Target

I think that we (including Eric Hansen) have been too ready to emphasize the 
non-equivalency or inequality between the equivalent and the equivalency 
target and have not taken sufficient notice of the equivalency and equality 
that are there.

We need to keep in mind that according to the definition of equivalent, 
there _is_ very important kind of symmetry between the equivalent and the 
equivalency target. That is, there is some "essential function" that is an 
intersection of the function served by the equivalent and that served by the 
equivalency target. As indicated in the definition of "equivalent", the 
_extent or size_ of this intersection between the respective functions of 
the equivalent and the equivalency target should be a large as possible and 
is limited only by the qualifying phrase "insofar as is feasible given that 
the state of technology and the nature of the disability".

Please note that regardless of the size or extent of this intersection, 
_there is symmetry between the equivalent and the equivalency target within 
realm of that intersection_! To push to point further, within the realm of 
that intersection, what we might term the 'feasible essential function', the 
equivalent and the equivalency target are not merely approximately the same 
but are equal in their capacity to provide function (e.g., meaning and 
benefit) to their respective audiences. Again, to emphasize, in this very 
important sense, the relationship between the equivalent and the equivalency 
target, there is symmetry.

I think that the only structural (i.e., necessary)  _asymmetry_ between 
equivalent and the equivalency target is that the equivalent has reference 
to a person with a disability and the equivalency target has reference to a 
person without a disability. And that asymmetry is there only because that 
is the only kind of equivalency relationship required by WCAG and 
necessitated by logic. There are plenty of other classes of equivalency that 
would have no such asymmetry.

SUGGESTIONS

Suggestion 1: Add an explanation to the definition of Equivalent that 
emphasizes the sameness of function and explains the rationale for the 
asymmetry that does exist.

Explain that reference to persons with and without disability is related 
directly to the objective promoting equal access. The first two sentences 
are the most critical.

Items 1 through 4 add additional clarifications. Item 4 mentions the 
potential for equivalency relationships that do not single out one 
equivalency object as the 'target'.

New content for the definition of Equivalent:

"The class of equivalency relationships that are the focus of this document 
is specifically intended to promote access by individuals with disabilities 
that is equal to that obtained by people without any disability. As such, 
these equivalency relationships assert an equality of essential function 
(insofar as is feasible given that the state of technology and the nature of 
the disability) between (a) persons with disabilities on the 'equivalent' 
end of the relationship and (b) persons without any disability on the 
'equivalency target' end of the relationship. It is important to note the 
following.
"(1) Due to inadequate markup and/or authoring, not all equivalency 
relationships may be recognizable.
"(2) The distinction between equivalent and equivalency target does not 
guarantee that the author of Web content intends either one for persons with 
a particular disability status. Markup language may allow an author to 
express intended audiences, yet the UAAG document attempts to make "all 
content" available to available to all users.
"(3) The existence of an equivalency relationship for content (e.g., content 
A equivalent for content B) does not exclude the possibility of another, 
reciprocal equivalency relationship (content B equivalent for content A) for 
the same content set.
"(4) There may be other classes of equivalency relationships that refer to 
equivalents but do not single out any of them as an equivalency target. For 
example, such classes might allow content C written in French and content D 
written in English to both to be considered 'equivalents' without 
designating one as the 'equivalency target'. Such equivalency relationships 
could be considered in designing effective languages and applications."

====

Suggestion 2: Consider the following new wording for checkpoint 2.3.

This new version makes a change suggested by Ian and addresses other issues 
in Note 2.

New (17 October 2000) Eric Hansen suggestion:

"2.3 Where multiple elements of content within a document are related by
_equivalency relationships_, provide easy access to each element
among the related elements. This shall be done through at least one of the 
following mechanisms: (1) allowing configuration to render one element 
instead of other related elements; (2) allowing configuration to render more 
than one of the related elements; (3) allowing the user to select a 
displayed element and then inspect any of the other elements; (4) providing 
a direct link to other related elements, just before or after a displayed 
element in document order. [Priority 1]"
"Note 1: For example, if an image in an HTML document has text equivalents, 
provide access to them by (1) allowing configuration to replace one element 
from among the image and its rendered equivalents;  (2) allowing 
configuration to render an image plus one or more of its rendered 
equivalents; (3) allowing the user to select a displayed image or equivalent 
and then inspect any of the other elements; (4) providing a direct link to 
other related elements (image or equivalent), just before or after a 
displayed element in document order."
"Note 2: While this checkpoint must include recognized equivalency pairs 
that are specifically cited in WCAG 1.0 (e.g., non-text elements and their 
text equivalents; multimedia presentations their auditory descriptions), 
user agent developers are encourage to consider providing the same 
functionality for other sets of content including: tables and their 
summaries and linearized versions; abbreviations and acronyms and their 
expansions; titles and respective objects; inaccessible pages and their 
alternatives. etc."
"Techniques for checkpoint 2.3"


Comment 1:

The main body of this checkpoint is the same as by 13 October suggestion, 
except that per Ian's suggestion I have deleted reference to the phrase "and 
similar" in the phrase "_equivalency and similar relationships_,". The other 
changes are in Note 2. The first Note has been numbered.

Comment 2:

Note that the scope is limited to those things declared by WCAG to be 
considered equivalents but people are encouraged to apply the principles 
more widely.

Comment 3:

Language translations are not mentioned in Note 2 but, as noted earlier, I 
have mentioned them in the point 4 of suggestion 1.

Comment 4:

Note that I have added "inaccessible pages and their accessible 
alternatives" in Note 2. Please keep in mind that Note 2 does not officially 
designate this or other optional pairs as equivalency relationships but 
encourages considering use of the same presentation and navigation 
approaches for them. For the record, I am using the terms "inaccessible 
pages" and their "alternatives" in an informal way. This terms have not yet, 
to my knowledge, received formal definitions.

Comment 5:

Note 2 refers to "sets" instead of pairs to account for many different 
equivalency 'objects' (see Suggestion 3).

==

For reference, here is my older version:

Old proposal (13 October) by Eric Hansen:

"2.3 Where multiple elements of content within a document are related by
_equivalency and similar relationships_, provide easy access to each element 
among the related elements. This shall be done through at least one of the 
following mechanisms: (1) allowing configuration to render one element 
instead of other related elements; (2) allowing configuration to render more 
than one of the related elements; (3) allowing the user to select a 
displayed element and then inspect any of the other elements; (4) providing 
a direct link to other related elements, just before or after a displayed 
element in document order.
[Priority 1]"
"Note: For example, if an image in an HTML document has text equivalents,
provide access to them by (1) allowing configuration to replace one element 
from among the image and its rendered equivalents;  (2) allowing
configuration to render an image plus one or more of its rendered
equivalents; (3) allowing the user to select a displayed image or equivalent 
and then inspect any of the other elements; (4) providing a direct link to 
other related elements (image or equivalent), just before or after a 
displayed element in document order."
"Techniques for checkpoint 2.3"

Also, add the following to the definition of Equivalent or in a Note:

"_Equivalency and similar relationships_ include at least the following:
equivalency relationships; linearizations and summaries of tables;
expansions (of acronyms); abbreviations (of phrases); language translations;
titles, etc."

==

For reference, here is the 29 September 2000 version:

Old (29 September 2000):

2.3 Provide easy access to each equivalent and each equivalency target
through at least one of the following mechanisms: (1) allowing configuration
to render the equivalent instead of the equivalency target; (2) allowing
configuration to render the equivalent in addition to the equivalency
target; (3) allowing the user to select the equivalency target and then
inspect its equivalents; (4) providing a direct link to the equivalent in
content, just before or after the equivalency target in document order.
[Priority 1]
Note: For example, if an image in an HTML document has text equivalents,
provide access to them (1) by replacing the image with the rendered
equivalents, (2) by rendering the equivalents near the image, (3) by
allowing select the image and then inspect its equivalents, or (4) by
allowing the user to follow readily available links to the equivalents.
Techniques for checkpoint 2.3

========

Suggestion 3: If necessary consider defining a new term -- 'equivalency 
object'.

If it becomes important to name an entity that more clearly puts the 
equivalent and the equivalency target on an equal or peer basis, then I 
might suggest the term "equivalency object". However, at this point, I see 
no real need for this new definition.

New:

"In the context of this document, an equivalency object consists of any of 
the objects joined by an equivalency relationship, such as the equivalent 
and the equivalency target."

========

Suggestion 4: Consider brief comments on Ian's proposal.

Below I have made some brief comments on Ian's 16 October 2000 proposals.

Part 1: Comments on Ian's responses
[3] http://lists.w3.org/Archives/Public/w3c-wai-ua/2000OctDec/0099.html

My 17 October comments are noted below with "EH:" Ian's words are noted with 
"IJ:".

IJ:

Hello,

Here's an attempt to summarize at least two issues being raised
and my views:

a) Issue 1: Can all equivalents be entirely symmetric?

    I believe the answer to this is no, and that means that
    the next question is moot: "Should WAI require that all
    equivalents be symmetric?"

    [For the purposes of this email, sets of content A, B, and C
    are "symmetric equivalents" if, with the appropriate
    environment and software, they all provide exactly the
    same browsing experience to some users. I suspect that there
    will be interesting commentary on this definition...]

EH: Here is one piece of such commentary. (Please note that this is 
essentially an aside from the main points of this memo.) As I think Ian 
anticipates, the phrase "same browsing experience" is problematic. By what 
criteria would one evaluate this "experience"? It apparently would not be 
the same _perceptual_ experience since one person might receive the 
'message' auditorily, another visually, and other tactually. It might not 
even be the same 'cognitive experience' in the sense that processes the 
different media might involve different parts of the brain and different 
cognitive processes. Perhaps then the sameness other experience should be 
evaluated by the sameness of "understanding". The definition of text 
equivalent uses the term "understanding" and this could be clarified even 
further, but there are no specific criteria for evaluating such sameness of 
understanding. (From my involvement in educational testing, I know that 
evaluating understanding can be no small task.) Another term that might be 
useful is "benefit" or perhaps "value". It seems clear that in the world of 
accessibility we are in the business of trying to achieve this sameness of 
benefit for _all users_ of information products and services.

IJ:

b) Issue 2: Should we use another word than "equivalent" to
    describe relationships that may not be symmetric for the
    user?

    No. Refer to proposal below.

ISSUE 1: Can all equivalents be entirely symmetric?

Al writes:
  "The order [among alternatives] is often present, but it
   comes in addition to the equivalency we care about,
   not as an intrinsic part."

I disagree. Listening to a symphony is not the same as reading
a musical score of that symphony (probably for everyone, including
gifted musicians, since there is interpretation in performance).
I don't believe that there's a "symmetric equivalent" for a
recorded or live symphony. Therefore, if I compose a symphony,
any attempt to provide an equivalent will necessarily result in
an imperfect equivalent (or alternative if you prefer to avoid
the term equivalent here). I think there is a lot
of content for which the author simply cannot provide a
browsing experience that is "exactly the same" for some
users (including some users with disabilities). Al writes:

   "there is some information that is represented in each
    [alternative] that is the same in both [/all]."
    This relationship is symmetrical.  This relationship is
    general. This relationship applies to an image and its alt text.

Yes, there will be "some information" captured in both pieces
of content, but not all information (due to inherent differences
between visual data and human language). So I disagree that
the relationship is symmetrical.

I propose that we should not attempt to apply a strict
mathematical framework when talking about "equivalence"
in WAI Guidelines. Or, please propose a definition
of "symmetric" that allows one to tell mathematically
that an image and its text equivalent are symmetric
in meaning.

EH: As noted earlier in this memo, this is or can be a high degree of 
symmetry with regard to "essential function". It is the focus on access 
audiences with different disability status that is the fundamental and 
necessary source of asymmetry.

IJ:

ISSUE 2: Should we use another word than "equivalent" to
    describe relationships that may not be symmetric?


2a) If we use the term "equivalent", will we confuse people?

    We may confuse some people, but if we explain that
    we are not using the term in its strict mathematical
    sense, I think we will be ok. As Eric has pointed out,
    the dictionary supports a non-mathematical usage.

2b) Should we continue to use the word "equivalent"
    (which is already used without a requirement
     for symmetry in WCAG 1.0)?

    I don't have particularly strong feelings about this,
    but I think we should keep the term at least until
    we have three Recommendations that use it in the
    same way.

PROPOSAL

1) State clearly in the definition of "equivalent" that
   we don't mean mathematical equivalence. We should state that:

   (a) where possible, equivalents should be symmetric
      (i.e., two sets of content should provide exactly
      the same experience to some users).

(b) symmetry is impossible in some cases, and ordered
       relationships will result: authors must provide
       "the closest you can get to equivalent".

EH: I think that the existing caveat "insofar as is feasible given that the 
state of technology and the nature of the disability" already covers this to 
the basic level and my new language further emphasizes this.

IJ:

2) Adopt most of Eric's proposed checkpoint 2.3 [1]. I think that Eric's
  "_equivalency and similar relationships_" introduces unnecessary
   ambiguity. I think it suffices to say in the definition of
   "equivalent" that the goal should be true symmetry.
  [1]
http://lists.w3.org/Archives/Public/w3c-wai-ua/2000OctDec/0092.html

EH: See my revised version that deletes the phrase "and similar".

IJ:

3) Delete "equivalency target" from the document (since only used
   in the glossary based on Eric's rewrite of 2.3).

EH: I am deeply concerned that following this suggestion will cause myself 
and many others to be drawn into a "tar pit" (terrible hindrance to 
progress) that the working group will have a hard time escaping. If we fail 
to properly name the thing that the equivalent is the equivalent for, then 
we leave a vacuum that will be filled by terms with undesired connotations. 
Furthermore the term "equivalency target" may be adopted for some other 
purpose, thereby compounding the confusion. I feel that to fail to solidify 
our gains and new clarity on the meaning of equivalent (as used in the 29 
September document) that these gains will be lost and we will be drawn back 
again and again to these same old issues. I urge that the term not be 
dropped unless a clear and acceptable alternative is found. I plan to 
register a minority opinion if there is a decision to return to the old term 
"primary content".

Please note that this memo suggests several specific changes that should 
reduce concern about the term "equivalency target". For example:

1. Point 4 of Suggestion 1 acknowledges that there may be other equivalency 
relationships that involve only "equivalents" rather designating an 
"equivalency target".
2. Point 2 of Suggestion 1 makes clear that even though "Markup language may 
allow an author to express intended audiences, yet the UAAG document 
attempts to make "all content" available to available to all users."
3. Point 3 of Suggestion 1 notes the possibility of reciprocal equivalency 
relationships for existing equivalency relationships.
4. Suggestion 3 suggests if it becomes important to name an entity that more 
clearly puts the equivalent and the equivalency target on an equal or peer 
basis, then the term "equivalency object".
5. Note 2 refers to "sets" instead of pairs to account for many different 
equivalency 'objects' in the same equivalency relationship.


CONCLUSION

In conclusion, I believe that these suggestions provide a great deal of 
latitude for expression of equivalency relationships. I hope that the reader 
will see that:

a. The symmetry and equality between the equivalent and the equivalency 
target is more important than we may have supposed.
b. The changes that I have proposed in this memo acknowledge and support 
equivalency relationships that are more strictly 'peer' in character than 
the equivalency relationship found in UAAG (29 September 2000) and WCAG 1.0.
c. The degree of asymmetry that is found between equivalent and equivalency 
target is the very minimum required by logic and adherence to the standard 
set by WCAG.


REFERENCES

[1] http://lists.w3.org/Archives/Public/w3c-wai-ua/2000OctDec/0092.html  
(Eric's 13 October memo)
[2] http://lists.w3.org/Archives/Public/w3c-wai-ua/2000OctDec/0094.html 
(Al's 16 October response to Eric)
[3] http://lists.w3.org/Archives/Public/w3c-wai-ua/2000OctDec/0099.html 
(Ian's 16 October proposed resolution)
[4] http://lists.w3.org/Archives/Public/w3c-wai-ua/2000OctDec/0100.html 
(Al's 16 October response to Ian)

<END OF MEMO>


_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.

Share information about yourself, create your own public profile at 
http://profiles.msn.com.

Received on Tuesday, 17 October 2000 12:55:26 UTC