validating microformats (was Re: [uf-discuss] Google Gdata new syndication protocol!)

[Luke Arno]

On 4/22/06, Tantek Çelik <tantek at cs.stanford.edu> wrote:
> On 4/21/06 7:18 PM, "Luke Arno" <luke.arno at gmail.com> wrote:
>
> > Mark P. is on the right track here as usual, I think.
>
> Yes.
>
> > What has been glossed over in this convo is that
> > underlying the discomfort that has been expressed
> > on the validation front here is this:
> >
> > Sure you can validate a microformat reasonably
> > well (though it may be difficult to validate format
> > transformations that result in microformats) but
> > there is not meta-microformat (god, did I just say
> > that?!?) to conduct validation automatically
> > (ie without manually translating the spec into
> > rules that your validator can understand) We
> > have profiles, of course but they are not machine
> > readable like a schema for tower of babel XML
> > (XSD or RNG or what not).
>
> Not glossed over.  I think you may have missed previous discussions which
> explained this quite simply.

Perhaps, I have been quite busy and can not follow
every thread. Forgive any gaps in my context.

Also "glossed over" was probably a poor choice of
words on my part as it implies some intentionality.
I should have just said that I hadn't noticed it stated.

>
> For *any* popular data format (e.g. HTML, RSS etc.), there is no meta-format
> that fully describes them, so the implied assumption that we should seek
> that goal for microformats is a poor assumption (or certainly one that is
> outside the scope of microformats).

I did not imply what you inferred: that we should have
such a thing. I thought that perhaps this was what
Bruce was getting at in the mail at the start of this
thread and wanted to clarify.

>
> In fact, DTD, Schema, etc. are insufficient to validate any real world
> adopted format, whether SGML, XML or something else.  Just go look at the
> source for validator.w3.org for HTML validation, or the source for the feed
> validator for RSS.  Any really useful XML will similarly need far more than
> DTD or schema to validate.
>

It is true that existing schema formats can only go
so far, but they express more to a machine than
a profile.

Besides these clarifications of fact, I only intended
to express 2 opinions:

1. I agree with Mark.
2. Schematron seems promising for uf validation.

Sorry for any confusion.

Cheers,
- Luke