Disambiguation [was RE: "uid" microformats? (was Re: [uf-discuss] ISBN mark-up)]

Wed Apr 26 11:53:55 PDT 2006

Thanks for the clarifications, especially to Tantek and Scott.  I've added a
bit to the FAQ. http://microformats.org/faq 

Tantek wrote:
>> Who is the registrar?
>Currently we have profile URLs at http://gmpg.org/  and http://w3.org/ 

Scott wrote:
>> Alternatively, who is to say which version of hCoupon is valid?
>Mark Pilgrim.  Rather, Mark is working on a validator, based on the  
specs at microformats.org.

>> Who is the registrar?
>http://gmpg.org/

It appears to me that there is no registrar.  Whether or not we need one is
a different issue, but I'd like to be clear in my understanding.  

>From reading through the material, the following appear to be true:
 1. There is no single place that one can go to dereference a microformat to
get a profile.
 2. There is no system to coordinate distributed dereferencing of
microformats. 
 3. There are no bookkeeping entities that serve as a disambiguators of
microformats.  
 4. There are no bookkeeping entities where one can register a microformat.

These are the four functions I would associate with a registrar. I should
clarify I mean an automated system.  Microformats.org as a social system and
human readable website certainly addresses some of these needs.  

Or perhaps I am misunderstanding some functions provided by http://gmpg.org/
or http://w3.org/?

This links back to the microformat UID/URL issue.  Does it make sense to
have a requirement or recommendation to specify the profile when specifying
the microformat?

Part of my challenge in understanding microformats and profiles is that all
the examples I came across for "vcard" simply used <div class="vcard">.

It seems to me that this creates a lot of problems. Especially with new
microformats.

(1) Some systems may not know that "vcard" or "hCoupon" is a microformat and
miss the data. How does any system know that a class is a microformat?

(2) When there are a lot of potential microformats, every system has to scan
every class attribute in every element against the entire dataset just to
check if it is in fact a microformat. This doesn't seem to scale.

(3) Page authors may not know about vcard or hCoupon but use the class
anyway.  This is especially likely with short descriptive names.

Given this, shouldn't the DIV be something like
<DIV class="microformat" microformat="hCoupon"
uid="guid:1F10E476-2E87-4c7e-AF2B-10DC40065764"
profile="http://www.joeandrieu.com/hCoupon"></DIV>

(Unfortunately this misses the elegance of the class name being the same as
the microformat name.)

Or because there will no doubt be future versions of the microformat
standard, we could reference the standard as well.

<DIV class="microformat" ref="http://www.microformats.org/v1"
microformat="hCoupon" uid="guid:1F10E476-2E87-4c7e-AF2B-10DC40065764"
profile="http://www.joeandrieu.com/hCoupon"> </DIV>

or perhaps something in the <HEAD> specifying which microformats are in use.

<meta name="microformat" content="profile:http://www.microformats.org/v1"/>
<meta name="microformatclass" content="class=hCoupon,
uid=guid:1F10E476-2E87-4c7e-AF2B-10DC40065764,
profile=http://www.joeandrieu.com/hCoupon/">

Then, in the page itself you can use <DIV class="hCoupon">.

I initially wrote up a number of other ways you might implement such a
scheme, but I guess first I'd like to understand if this challenge is one
that is either (a) irrelevant/unimportant or (b) already addressed by
something in microformats I don't understand yet.

This thinking is leading me to slowly approximate the DOCTYPE tag from XML,
but hopefully in a way that fits within the structure/model of microformats.

It just doesn't seem to me that <div class="vcard"> scales to a wide variety
of microformats nor to discontinuities in the microformat specification
itself (or in individual microformats for that matter).  It looks like it
works for well-defined, formally understood terms, but not for innovative
new uses.  But again, I worry I'm spinning my wheels here.

Am I missing something or is disambiguation an unresolved problem?

-j

--
Joe Andrieu
joe at andrieu.net
+1 (805) 705-8651