Re: img@relaxed CP [was: CfC: Close ISSUE-206: meta-generator by Amicable Resolution]

John Foliot writes:

> On August 01, 2012 Henri Sivonen wrote:
> > Do you really think this is about making things nicer to validator
> > users?
> But then, On Aug 2, 2012 Henri Sivonen wrote:
> > I believe the key thing is whether the markup generator developer
> > thinks that the validator output reflects badly on the quality of
> > the generator in the eyes of ill-informed people who use validation
> > as a quick smoketest of generator quality. I expect any suggestion
> > that the output of the generator is "incomplete" to trigger the
> > worry that the validator output reflects badly on the quality of the
> > generator in the eyes of ill-informed people who use validation is a
> > quick smoketest to such a degree that some markup generator
> > developers want to make said validator output go away.
> Let's see here... In 2 sentences I count 6 instances of "generator", 3
> instances of "validator output", and 2 instances of "validation" (and
> zero mention of end user), so ya, I believe that you are primarily
> interested in what the validator outputs,

I can't speak for Henri, but I'm interested in what the validator
outputs _so_ as to best help end users.

Validation for its own sake is pointless.

Validation only has utility where its output causes a developer to
change the HTML being validated, and where that change is better for end

So we (certainly, I) want for the validator to yield the output which
results in the best chance of nudging developers to make changes that
are helpful for end users.

> to the point where the end user's needs are being trumped by your
> concern. 

I don't see how that follows from the Henri's text above. There's a
chain of spec text -> validator behaviour -> validator output messages
-> developer interpretation of that output -> developer changes -> what
HTML gets served to an end user. That a particular paragraph happens to
mention one part of that chain more than another doesn't mean that the
end result isn't important.

Quite to the contrary, we need for all parts of that chain to be
considered in order to have the best possible end result.

It is of course possible that there are others in this thread whose
motives are different from mine; but none of what Henri wrote is
inconsistent with somebody earnestly trying to produce the best output
for end users, so please can we proceed on that basis, and consider the
substance of the point he made, rather than attacking potential motives
behind it?

As I mentioned a few days ago, this mailing list has a history of
seeming more hostile than other technical lists I read, and I for one
find the hostility off-putting, pushing me away from wanting to be
involved in improving HTML.

I don't think it's good for HTML to be designed only by those with
robust constitutions; surely the views of timid people should also be

Having seen somebody's motives attacked on the list like this, it makes
me reluctant to share my own thoughts with the list, lest my own motives
be similarly attacked.

> You want the validator results to tell the generator tool and those
> "ill-informed people" using such tools and who, surprise surprise,
> want a validator to tell them if they are doing things right, that
> bravo, you're doing a good job (even if it/they aren't). How else can
> anyone interpret your statements?

Again I'm not Henri, but this is what I want when an developer uses a
conformance checker on some HTML:

  A For any situation in which the developer can improve the HTML, for
    the validator to give output which nudges the developer into making
    those improvements.

  B For any situation in which the developer can't improve the HTML (so
    far as the validator can tell), for the validator to give output
    which indicates that -- and in particular for it _not_ to give
    output which (inadvertently) nudges the developer into making a
    change which makes things worse for users.

I didn't think that was controversial; so far as I can tell, that's
pretty much the whole purpose of validators.

But given the gap between different opinions on this thread, and the
potential for misunderstandings, I think it's worth checking, to help to
narrow down the precise matter there's disagreement on.

So, for the avoidance of doubt: John, are the above outcomes A and B
also what you want from a developer using a validator?

If not, please can you explain what it is that you would like to happen,
if that's possible without attacking my motives or questioning my

And everybody else reading this: if you want a validator to do something
different than A and B above, please can you explain what it is.


If the above does match your views, then it seems to me that the nub of
the issue here is the following 2 questions. These are both questions to
which it seems there are differing opinions on the list as to the best
answers, and on which I haven't yet personally formed an opinion:

  1 In the situation where the developer responsible for some HTML that
    includes an image does not know the image's content and purpose at
    the time she's writing the HTML, and where the image is added by a
    later non-interactive batch process -- with no opportunity for the
    provider of the images to be prompted for suitable alt text
    -- what is the best HTML for users who aren't currently seeing
    images (whether that be because they are blind, or using Lynx, or
    for any other reason)?

  2 Given the answer to question 1, what is the most useful validator
    output to maximize the outcomes A and B above?

    That is, what's the best validator output for HTML in the situation
    described in question 1 such that sub-optimal HTML is improved and
    HTML that can't be improved any more by the developer isn't made

Since there are people on this list who I believe do have views on the
answers to those questions, and since I haven't formed one myself, I'm
hoping that those with such views could share them and the reasoning
behind them, with the aim of persuading me (and others on the list) to
adopt their position.

They strike me as questions on which reasonable people can disagree. I
would hope that people could disagree on the substance of the questions
without thinking ill of each others' motives.

Apart from its off-puttingness, the hostility doesn't actually assist in
any way with helping form opinions on the questions above.

> On  August 02, 2012 Henri Sivonen wrote:
> > alerting users is what triggers the behavior in (some) markup
> > generator developers that we're trying to avoid triggering! I expect
> > that the kind of the markup generator developers who wanted to
> > silence HTML4 validators with alt="" would want to make this warning
> > go away, too, so I expect this proposal would not yield a worthwhile
> > benefit compared to going back to the HTML4 situation
> And you should be honest enough to admit that validators, and more
> importantly junk generators that don't even offer the opportunity to
> do the proper *and truly conformant* thing, are gaining way more in
> this emergent compromise than non-sighted users.

Again I find that accusation of dishonesty to be fostering a hostile
environment on this list, as well as being unnecessary to actually
discussing the nub of this issue.

Best wishes


Received on Friday, 3 August 2012 09:09:48 UTC