[uf-discuss] uF Discovery?

Luke Arno luke.arno at gmail.com
Thu Nov 24 10:32:13 PST 2005


On 11/24/05, Brian Suda <brian.suda at gmail.com> wrote:
> I'm not avocating something for discivery MUST be built. I am saying
> "lets look at how other guys have tried to figure this out, has it
> worked, or failed, and what can we learn from that?"
>

Right on.

> HTML has been successfully broken down into several pieces, CSS for
> presentation, HTML for data, and Javascript as yet another layer.
>
> Microformats use HTML as the data layer, XMDP as the decription layer,
> but there is no discovery (maybe there doesn't need to be?).
>

I think we can really benefit from some discovery
mechanism rather than being limited to the blind
crawl.

> As Luke pointed out:
> > Software that uses distributed uF data may want to
> > see where all the uF data is and not get caught up
> > in crawling the whole web. Context: REST APIs.
> >
> > - Luke
>
> the advantage of saying what you have available will minumized the
> crawl space. I can get one file that tells me everything, or crawl the
> entire 40,000 Avon hCard pages to try to get the same thing.
>

Or just links to full cards that match a search.

> Look at Google Sitemaps, that is a single file that describes the
> pages on the site, along with last-update time. This helps to limit
> un-needed crawls, bandwidth, time, etc.
>
> This aproach has its downsides as well. Data drift is my biggest concern.
>

Drift is always going to be an issue. "Cool URLs don't
change."

What I am after is a way to represent distributed
datasets (paginated listings and various related
items) with uFs.  A REST style service will often
provide some listing (hopefully XOXO):

GET /contacts

...
<li>
  <a href="/larry#hCard" title="Larry" class="disco-hCard">Larry</a>
</li>
<li>
  <a href="/moe#hCard" title="Moe" class="disco-hCard">Moe</a>
</li>
...

Follow a link in the listing to get a full editable
hCard.

GET /larry#hCard

[ Larry's hCard ]

Parse and edit larries info then PUT it back.

PUT /larry/#hCard

[ Larry's modified hCard ]

Sorry for getting so hypothetical. I promise to
write a formal problem statement and start
gathering examples strait away, if there is
interest. :-)

Does anyone else feel that this is worth getting
started on the wiki?

- Luke


More information about the microformats-discuss mailing list