Re: Working Group Decision on ISSUE-31 / ISSUE-80 validation survey

On Thu, Apr 21, 2011 at 9:59 PM, John Foliot <jfoliot@stanford.edu> wrote:
> Not wanting to be confrontational, is this statement based upon any kind of
> formal measurement or assessment, or rather via anecdotal comments and
> observations?

I'm speaking only from my personal perspective as a web developer.  I
can only speak for what I feel myself and what I've observed among
other web developers I know.  (I'll specifically point out that
although I happen to be working on a contract for Google right now,
what I'm saying in this thread is on my own time and unrelated to my
affiliation with Google.)

> I also ask, because just recently Jonas Sicking stated:
>
>        "A terrifyingly small percentage of the pages on the web pass a validator.
> The far vast majority of pages doesn't even nest their tags correctly. The
> sad truth is that while we can do what you suggest, it's not going to have a
> big effect because people simply doesn't consult validators to a large
> degree."
> http://lists.w3.org/Archives/Public/public-html/2011Mar/0627.html

Jonas is quite correct, and it doesn't contradict what I say.  All of
the following things are true (in my experience):

* The large majority of web authors and users don't know or care about
standards or validators at all.

* Among particularly savvy web authors, there's much more awareness
about standards, and people do care somewhat about standards
conformance and do use validators, but usually not beyond the point of
just getting the validator to say their page is okay (if that).
However, these web authors are often disproportionately influential,
like writing high-profile web software or web development tutorials or
books.

* When deciding whether to purchase or use particular web software,
the large majority of users ("users" meaning "people who will install
the software on their site", not "end users of the software") don't
care at all about standards, but a small but significant minority
does, such that it's often worthwhile for software authors to make
sure their pages validate so they can claim standards compliance.

> As an invested participant, I am now hearing conflicting commentaries from
> others involved in the discussion, so it would be useful if we all knew what
> the real status is/was: validation *does* matter, or *doesn't* matter. I
> pose this question to the larger Working Group as well.

Validation matters to a small but potentially significant percentage
of authors.  Conforming to the actual intent or word of the standards,
beyond merely achieving validation or perhaps achieving a few other
goals that are circulated by word of mouth ("avoid layout tables!"),
matters to a vanishingly small and basically insignificant percentage
of authors.  So insofar as standards authors can have any effect on
web developers, it's mostly by deciding what's valid markup.  (In my
experience.)

> I am happy to see you pluralize "standards", as this is indeed an important
> point to consider.  Within the W3C, we currently have a Standard, WCAG 2.0
> (W3C Recommendation 11 December 2008) that specifically states:
>
>        "1.1.1 Non-text Content: All non-text content that is presented to the user
> has a text alternative that serves the equivalent purpose, except for the
> situations listed below. (Level A)" -
> http://www.w3.org/TR/WCAG20/#text-equiv

For the use-case of user-uploaded images, this requirement is simply
impossible.  I, as a web developer, cannot make my software output
text equivalents for things that users have uploaded, because I don't
know in advance what they are or how they're being used.  (Even if I
prompted users for alt text, I still couldn't ensure that they put in
text that actually serves an equivalent purpose to the image.)

> Would it not then be fair to say that using a mechanism that seeks to defeat
> or obscure conformance to these Standards could also result in "... some
> percentage of users hold(ing) that against it, and assume that it reflects
> some deficiency in the software (whether or not they actually understand the
> errors)"?

It could, but in my experience it doesn't, or at least not as much.
I've seen more than a few complaints about MediaWiki pages not
validating, like here:

https://bugzilla.wikimedia.org/show_bug.cgi?id=24500

Notice how there were three duplicate bugs, so at least four separate
people filed bugs on the issue, even though it has absolutely no
visible effect and is even valid in HTML5.  There are lots and lots
more bugs I've seen filed like this, and I've seen such complaints
against vBulletin too.  In almost all cases, users only report
validation errors, not other conformance errors.

I've seen a couple of bugs filed by the users of screen readers about
various problems in MediaWiki, but only a couple.  I've never seen a
bug filed about our current image alt handling for cases where no alt
text is specified, and a cursory search suggests that there is no such
bug that anyone has filed.  So it's not something that seems to have
much user demand, relative to other bugs and feature requests.

> I wonder aloud if representatives of companies such as Microsoft, Apple,
> IBM, Oracle, etc. would concur with that statement. I wonder too how Google
> feels about this position today with regard to their already significant
> investment into Googledocs, and the recent problems they are facing there?

I have no idea.  I can only speak from my own experience.  Clearly
some companies invest significant resources in some types of
accessibility.  However, I'm still fairly sure that *most* companies
do not include disabled users in routine testing or QA (keeping in
mind that most companies are quite small).  Undoubtedly some do,
especially for extremely large, widely-used, extensively-tested
products like an operating system, but a typical smaller product has
little enough formal testing of any kind.  Take that assessment for
whatever you feel it's worth.

I reiterate that I am not speaking for Google here in any way.  I'm
speaking mostly from my experience as a volunteer MediaWiki developer,
partly from my experience following vBulletin development as a user,
partly from what I've seen in general.  In fact, I have no regular
contact with anyone at Google and don't consult anyone else on what I
post here at all, as a rule, so I'm never speaking for Google unless I
explicitly say I am (which I never have, and likely never will as a
contractor).

> As some-one who has worked in the web-accessibility field for over a decade,
> I have often heard these types of claims. For sure, this was a claim that
> Target.com made early on in their legal woes, and yet, once the dust settled
> (and Target settled out of court) the final tally for that adventure was a
> settlement that made changes to their Web site (cost undisclosed, but
> generously I will suggest $1 million dollars - that surely must buy a lot of
> changes!) and further set aside *$6 million* for plaintiffs to share.
> Target's failure to include disabled users in testing, planning, or QA here
> surely must have had a significant impact on their ROI after the fact.

Well, okay, if you risk being sued, that changes things a bit.  I
don't know about that kind of thing, I'm not a lawyer or a
businessman.  I'm pretty sure it doesn't affect the software I
develop, which is open-source and distributed free of charge.  So I'm
just focusing on what requests I see from users.

> Aryeh, again in my personal experience, when explaining what it is I do,
> even the most naïve and uninformed person instantly understands that images
> that cannot be seen (by blind users - and we don't need to discuss all the
> other scenarios, blind users here suffice) require textual alternatives - it
> cannot get more basic than that.

That's all very well, but we were talking about what alt text software
should generate for user-uploaded images when the user did not provide
any alt text or caption.  HTML5 says the software should either omit
the alt attribute or use alt="", while WCAG and "HTML5: Techniques for
providing useful text alternatives" give no guidance.  The fact that
blind users cannot see images is pretty obvious, but it doesn't help
with this question.

> I agree. It clearly is primarily an educational effort, as you likely know
> that providing a means to associate alternative text to images inside of
> MediWiki's data base would be rather trivial to implement.

I did implement it, in October 2008:

https://bugzilla.wikimedia.org/show_bug.cgi?id=368
http://www.mediawiki.org/wiki/Special:Code/MediaWiki/41837

However, MediaWiki still needs to generate some default alt text in
cases where the user didn't provide any, which is what we were
discussing.

> It's not the
> software where the problem lies, it's the authors, who through unwillingness
> or lack-of-awareness produce incomplete content when they omit textual
> alternatives with their images.

In the case of the big MediaWiki installations like Wikipedia, the
authors include potentially everyone on the planet, since anyone can
edit a wiki.  Suggesting that all the authors of Wikipedia content
should be educated to provide appropriate alt text is a non-starter.
We're left with the question of what to do when they don't provide any
alt text.

> As some-one who struggles with the 'educational' piece almost daily (the
> largest part of my day-to-day job at Stanford is focused on author training
> and awareness), I have come to very much appreciate that getting good
> outcomes is closely tied to good work-flow processes. If, (for example), I
> could not upload an image to MediWiki without providing a textual
> alternative to that image, then what I would have would be a work-flow that
> makes doing the right thing easier, and the wrong thing harder.

It would also impose a burden on 99% of users for the benefit of 1%
(figures made up, I don't know the actual numbers).  Plus there's no
reason to believe it will actually produce useful alt text.  So I
don't think MediaWiki or any other major web software is going to do
this anytime soon.  Someone filed a bug asking for this in 2005, and
another MediaWiki developer closed it WONTFIX last year, apparently
after discussing it with other developers (not me):

https://bugzilla.wikimedia.org/show_bug.cgi?id=2443#c15

Viewed another way: no software that I've heard of that allows
user-uploaded images requires uploaders to provide alt text.  Given
that, guidance is needed in such a scenario, unless you expect all
such software to change.  HTML5 is the only standard that seems to
even try providing guidance, and I'm not sure it's sound.

> And while I have to admit that creating this type of work-flow inside of
> MediaWiki is likely significantly harder than adding a "generator" string to
> the tool and walking away secure in the knowledge that pages will now
> validate, it just feels and appears oh-so-wrong to do. It kind of reminds me
> of drivers with diplomatic license plates parking in handicap spots with
> disregard, because they are exempt from parking tickets.
>
> Is that really what the collective *we* want from the web? Please say no.

Developers don't have dictatorial control over our software's
features, unless we don't care if anyone uses it.  You can't get away
with adding features that annoy the vast majority of users for the
benefit of a small minority.  Users will complain loudly, and if
they're ignored, many will either disable the feature somehow, or stop
using the software.  Too many features like this, and your software
will lose out to competitors.  Developers can only pressure their
users at high cost.  If you want to force users to do something for
the greater good even though they don't want to do it, you're going to
have to speak to legislators, not software developers.

> If you really wanted emails from blind users, I could probably manage to
> have your inbox explode with those requests - I am pretty good at shaking
> the trees when I need to. I won't do that however, as spamming you would be
> counter-productive, and I really want to not be counter-productive here.

It would also not help your point.  User requests are only a
reasonable proxy for what users actually want as long as they're
spontaneous.  If you organize a whole bunch of people to sign a
petition or spam a bug tracker or whatever, that doesn't prove that
the feature is wanted any more than other features, it just proves
that one person cares enough to round up signers for a petition or bug
tracker spam or whatever.

My observation is that *of their own accord*, users file many
complaints about validation errors (for instance) but very few about
accessibility.  That suggests that many more users care about
validation issues with MediaWiki than accessibility issues with
MediaWiki.  (Which could be because MediaWiki's accessibility is good
enough that users don't feel the need to file bugs.)

> For the most part MediaWiki is a pretty good tool. But have you ever asked
> end users, including users with disabilities, how you could make it better?

I haven't, especially.  The Wikimedia Foundation has funded strategic
initiatives and usability studies.  At least a few users appear to
have suggested that Wikimedia prioritize accessibility in some way.  I
could find three out of some hundreds of proposals on
strategy.wikimedia.org:

http://strategy.wikimedia.org/wiki/Proposal:Accessibility_initiative
http://strategy.wikimedia.org/wiki/Proposal:Create_an_accessibility_committee
http://strategy.wikimedia.org/wiki/Proposal:Accessibility_for_deaf_and_visually_impaired_communities

None of them seem to have garnered much attention (like almost all the
proposals there).

> As to what kind of change to make? I offered one suggestion, but why not ask
> blind users what they think? Or mainstream, "I just want it to work" users?
> I would be happy to help you conduct a managed survey of those kinds of
> questions, if you and the MediaWiki team were truly interested in knowing
> those answers - I can access a decent pool of either type of user. Ping me
> off-line, and I will make that happen for you - that's a promise you can
> take to the bank!

I'm really not active in MediaWiki development right now, so I'm not
the person to ask.  You could try sending an e-mail to wikitech-l:

https://lists.wikimedia.org/mailman/listinfo/wikitech-l

> In practice today, images with null alt values are effectively silenced.
> (Actually, the screen reader essentially "reads" *null*, thus nothing is
> announced). In the absence of alt text (and the alt attribute) screen
> readers attempt to do something, and so announce the file name, reading
> aloud such gems as "one seven nine zero five five four seven one dot
> jay-peg" - so when the 'current HTML spec" says to just omit the alt
> attribute you are causing real grief and pain to screen reader users. Don't
> do it!

Actually, MediaWiki files tend to have pretty comprehensible
filenames, like "294px-Nafaanra_language.svg.png".  Files with numeric
names tend to be renamed quickly, since wikitext syntax requires
referring to files by name.  The software could provide the filename
with extensions stripped as default alt text, like in the case of this
image "Nafaanra language", if that was better than the empty string.
It could also be added only if there's no caption.  But most
uncaptioned images on Wikipedia are decorative, so maybe the status
quo is fine.  All I know is, no one's complained that I've heard of.

Received on Friday, 22 April 2011 19:30:16 UTC