Re: [XML] well-formedness & validation / section 4.2.8.1

Eduardo Casais wrote:
>> I add that a "SHOULD" statement is already strong, 
>> especially since we will require conformant deployments
>> to justify the reasons for not  following a SHOULD
>> statement.
> 
> Two comments on this:
> a) This fails to address point (3) in my message. Once one has found a justification to escape the validation clause, there is not even a minimum guarantee of well-formedness.

Right. No minimum "guarantee".
That is where I think that a SHOULD binding is strong. According to 
RFC2119, "the full implications must be understood and carefully weighed 
before choosing a different course". It is strong-enough in my view in 
that it says validation (and thus well-formedness for XML content) is to 
respected, and that any deviation to validation (and thus 
well-formedness for XML content) is to be given much care.


> b) According to the reasoning, this is actually a justification to eliminate entirely the reference on validation. I do not quite see why one would impose a supposedly strong requirement on validation, while at the same time arguing that well-formedness, which is formally much weaker than validation, is too strong... There is an inherent contradiction here.

I do not understand the contradiction. I am saying well-formedness and 
validation are equally strong.


>> Mandating well-formedness is pretty cool, but I suspect 
>> well-formed content is still the exception to the rule on 
>> the Web, especially with  legacy Web sites (I 
>> understand that the mobile Web is by far "cleaner" in 
>> that respect than the old desktop one, but that is not the
>> point here).
> 
> The point is as follows: insofar as the intent is to make legacy, non necessarily valid or well-formed desktop content available to mobile devices, one has to consider the target formats these support:
> a) XHTML mobile profile: XML dialect, requires well-formedness.
> b) XHTML basic: XML dialect, requires well-formedness.
> c) WML: XML dialect, requires well-formedness.
> d) HTML: does not a formal definition of well-formedness, only of validation.
> 
> Conclusion: either one must enforce well-formedness, or one cannot because the concept does not exist.

I follow your point, just does not see why it is so important. What I 
mean is that I do not think that any CT-proxy wants to produce content 
that breaks the rendering on mobile phones, and is likely to take care 
of that by itself.


>> What we could say is "When the initial content is well-
>> formed, the altered content MUST be well-formed".
> 
> The assumption is that the CT-proxy will modify the input content "in-place" and not convert it to another format -- which I suspect actually represents the majority of the cases (see above).

Well, it would also work for conversion between some flavor of XHTML to 
XHTML Basic. A recent study from Opera seems to indicate that there are 
quite a few XHTML "transitional" pages on the Web:
  http://dev.opera.com/articles/view/mama-basic-document-structure/#doctypes
... but I am more suggesting that we stick to "SHOULD validate according 
to an appropriate published grammar" statement anyway.

Francois.

Received on Monday, 1 December 2008 14:08:57 UTC