Re: Understanding the "applicable specifications" clause (was: Re: Decentralised extensibility idea (ISSUE-41))

Henri Sivonen, Wed, 20 Jan 2010 16:23:57 +0200:
> On Jan 20, 2010, at 15:50, Leif Halvard Silli wrote:
>> Henri Sivonen, Wed, 20 Jan 2010 14:56:30 +0200:
>>> On Jan 20, 2010, at 13:50, Leif Halvard Silli wrote:
>>>
>>>> What most people will be interested in is what the W3 Validator 
>>>> considers valid.
>>> 
>>> Indeed, it probably a more relevant consideration for many than what 
>>> the specs say.
>> 
>> I for one agree with you that a validator should be a 
>> development/authoring tool. 
>> 
>> However, with the "applicable specification" concept as the only 
>> extension point,
> 
> It's not the only extension point.

It would be interesting if you can point to other _specification_ 
extension points which also results in documents that get a valid 
stamp. 

>> the weight of what the (W3) validator considers valid 
>> will only increase in importance against the weight of the spec.
> 
> I don't see how it would increase compared to how things have been 
> for quite some time with the W3C Validator.

So it is it your view that the W3C validator has accepted more than 
HTML4 accepts? It is probably right w.r.t to IRI. The W3C validator has 
also not validated things that are within HTML4 but which requires - or 
should have had - a @profile in order to be valid: extended semantics 
@rel/@rev/<meta>.

Anything else? The W3C validator does not accept that you use 
attributes that are not part of HTML4. Whereas the "applicable 
specification" concept doesn't hinder that whether elements nor 
attributes are added. That is a big difference. 

>> One way to make the point that "validation is not all that matters" is 
>> to make it very simple to validate stuff (= easy to add extensions that 
>> can be validated with "off the shelf" validators = having technical 
>> extensions rather than "applicable" extensions.).
>> 
>> If you on purpose want to make it difficult for people to validate 
>> stuff that is not part of the main spec, then you are not operating 
>> with a pure development view of what validation is meant for. Instead 
>> you are mixing in your perception of how (for example) I perceive 
>> validators/validation.
> 
> Is that a personal "you" at me or an impersonal English "you"?

I meant: a policy which says that "applicable specification" is the 
only _specification_ extension point makes it difficult to use a 
validator as a development tool. When I say "development tool" then I 
still meant the W3 validator - not some private validator.

My primary point is about @profile: I don't expect that validators 
should be able to know - and thus perform validation - of all possible 
profiles, but I expect that they should not consider that I have made a 
suspicious document just because I used @profile. When I use @profile, 
then I am open about the semantics I use. I wonder why that should be 
punished.

> In case it is a personal "you":
> Validator.nu allows its built-in features to be replaced piecewise 
> with custom schemas. That is, you could feed Validator.nu your own 
> RELAX NG schema and still use the built-in table integrity checker, 
> for example. The particular validator extension mechanisms supported 
> by Validator.nu at a given point in time aren't (and, in my opinion 
> shouldn't be) coupled with the spec extension mechanisms of the HTML5 
> spec.

It is of course nice if I can do this offline on my own computer.  But 
if I want to validate a document which does use an attribute or element 
that the HTML5 language specification doesn't contain, using the W3 
validator, then can I do it automatically? Can others do it 
automatically?

(Again, I care most about @profile, from HTML4. I am fine with it if 
new elements "require more" to be accepted by validators.)
 
>>>> b) if you do go for an 
>>>> extension, make sure that it can be W3 validated = best thing is to 
>>>> submit the spec to the W3.
>>> 
>>> That probably helps. However, it seems that it's neither necessary 
>>> (Atom validation is offered)
>> 
>> <feed xmlns="http://www.w3.org/2005/Atom">
>> 
>> I've read the Atom history. But Atom has both a W3 namespace and is 
>> based on XML.
> 
> The W3C also provides RSS 2.0 validation, and RSS 2.0 doesn't even 
> have a w3.org namespace URL. RSS 2.0 isn't in any sense a W3C spec.

It at least purports to be XML. According to the pundits there is only 
two kinds of mark-up online - both of them stems from W3C: HTML and XML.

>> That way it can also be be validated as part of an XHTML 
>> document, I suppose.
> 
> I meant http://validator.w3.org/feed/. I didn't mean Atom inside 
> XHTML (which would be weird).

But w.r.t. HTML5, then we want "weird": We are talking about extending 
what is permitted inside HTML5 documents.
 
>>> nor sufficient (XForms validation is not  offered).
>> 
>> It is also a significantly newer recommendation.
> 
> XForms 1.0 became a REC in 2003. The development of Atom started in 2003.

OK. I don't know, but I guess this so because the people behind 
Validator.w3.org didn't find XForms to be valid on the Web? But what 
does this prove?
-- 
leif halvard silli

Received on Wednesday, 20 January 2010 17:11:26 UTC