Helio Miranda <heliomx@g...> writes:

> ... I've extended some XHTML tags to add an form
> validation functionality in the client side (implemented with javascript
> ). I've created a set of attributes that indicates what kind of
> validation check is needed for the form field before it was submitted to
> the server. Like this:
>
> <input type="text" v:type="email" />
> <input type="text" v:type="name" />
>
> Then a javascript function processes all the tags and validates the form
> accordingly to the v:type attribute. 
>
> Everything was fine. All the modern desktop browsers supported the
> extension, but that's not enough for the project I've been working on.
> The code must be validated by some XHTML validators around the web. The
> code validation is required for "quality of the product" reasons. These
> validators were complaining about the invalid/unspecified attribute used
> in the <input /> tag.
>
> So I've created a Schema for the "v" namespace. The problem is: I've
> tried to extend the XHTML specification and turn the code into a valid
> XML document (the input tag must be expecting a possible v:type
> attribute occurrence) . I found it was a hard task to conclude.
>
> So, can you give me suggestions of a painless way for doing it? And what
> validation engines support such schema extension?

Painless?  I doubt it - judging by your description, pain
is part of the structure of the situation.

HTML validators that use the standard HTML DTDs are unlikely 
to aceept your attribute, because DTDs have no construct for
saying "any attribute in any namespace other than the HTML
namespace is OK here".  The HTML DTDs are heavily parameterized,
so in theory you may be able to define a parameter entity that
makes the attribute v:type legal in the input element.  But
it won't be much fun.

(X)HTML validators that use XSD schemas to validate will, or
won't, accept your attribute as legal, depending on whether the
schema they are using does or does not have an attribute wildcard
on the input element.  There are several sets of XSD
schema documents for XHTML on the Web, some of which are
attempting to mimic the behavior of the DTDs as exactly as
possible (so they don't have attribute wildcards), and
some of which are attempting to match the spec more
closely (so they do).  Schemas with the latter should
accept your v:type attribute whether you provide a
schema for your namespace or not (because the wildcards
in question are lax).

Whether a validation service supports extensions or
modifications to the schema being used depends on the
purpose and policy of the service.  If the point is to
provide a running validator, so that users of the service
don't need to install one, then it's plausible to 
allow the user to specify what schema is to be used.
But if the point is to provide some more or less
objective test of the conformance of a document to
a specified schema, then the service won't typically
want to use the schema you provide -- it will want to
use the schema it already knows about from trusted
sources.  You aren't a trusted source - if you were,
your document wouldn't need validating, would it?  
So I wouldn't expect most HTML validators to allow you
to specify your own copy of the relevant schema
documents. 

Whether that puts you irrevocably at loggerheads with those
who want validation of the document, I can't tell.

If you absolutely must make the HTML valid against
the DTD, or against a schema that doesn't allow for
out-of-namespace attributes, then I can think of 
one possible way out:  use the built-in semantic
extension provided by the 'class' attribute.  So instead
of v:type="email", you write class="email".  It's
not even tag abuse -- that kind of thing is (as far as
I understand) what class was introduced for.

I hope this helps.

--C. M. Sperberg-McQueen
  World Wide Web Consortium