[rdfweb-dev] Catching up and moving forward

Dan Brickley danbri at w3.org
Wed Jul 23 16:14:01 UTC 2003

* Marc Canter <marc at broadbandmechanics.com> [2003-07-23 08:31-0700]
> First of all - thank you to Morten for pointing me at the archives and
> to Shelly for pointing out. well let's just say I have a lot to learn to
> play here.

re archives, as of pretty recently, our email archives are 
slightly easier to find (RSS feeds too) from the 'developer' 
sidebar of http://rdfweb.org/  I'm not that happy with the default 
MailMan web archives we have at moment, no searching, threading is 
crap, etc etc., but as out of the box stuff goes it seems bearable. 
The RSS feed is an add-on, give me a shout if it goes wrong, we're 
running my patch of someone else's patch of the MailMan tools. 
Thanks to an 'extract yourself from YahooGroups'
utility, the archive goes back to June 2000. The todo list I posted back
then looks worryingly familiar ;)
> 1.	We're in sync with version numbering, location of spec, who's in
> charge and where this is all gonna happen.  Our tool will have a formal
> staging server - so as you gals and guys iterate and evolve - we'll test
> first, then deploy - assuming that we're in a dynamic petrie dish kind
> of world.

sounds like a plan!

> 2.	I also think I understand the duality/RDF way of things. I just
> hope Eric Sigler (one of our programmers) can make a decent
> parser/interpreter that can handle all variations. I think that's
> crucial.  But I'm confidant Eric is up to the challenge!

Eric's done some nice things with FOAF before, so should be fine. The
way people normally build RDF applications is to use an 'RDF parser'
which is a layer that sits between your code and the XML parser (Expat
or whatever) that sees the angle brackets. It is the business of the RDF 
parser to know about the various different XML encodings of RDF triples, 
so your typical application writer shouldn't have to write code that 
cares about purely syntactic variations, any more than they should be 
caring whether XML files use ' or " to enclose attribute values. I think 
Eric is familiar with the XML::FOAF Perl module and the RDF::Core
library which it uses for parser etc., so this should be pretty
straightforward for him.

> 3.	The issue of Slashdotting - is also key. Has anyone ever
> approached Brewester Kahle and the Internet Archive?  Infinite bandwidth
> and storage are available for projects like what you folks are up to.
> We'll certainly be hosting our stuff there - once it's deployed.

I think Bill Kearney has contacts there, I've not looked into what they 
might offer. I think a useful goal for those of use with FOAF-aggregating
sites is for at least some of them be relatively slashdot proof, through 
traditional means of caching, static files vs db-lookups, etc. And to 
accompany that with downloadable code so others can pick up and run 
their own harvesters.

> 4.	I'm confused over foaf:group?

Join the club ;)

There are a couple of similar goals that get easily mixed up in

(i) there are social entities we call groups, where members of the 
group treat the group as a 'thing' and would acknowledge its existence
in some sense. 

(ii) there are infinitely many ways of grouping people based on their 
characteristics - eg. single white females living in NYC; unemployed 
programmers with an interest in politics; people with myers briggs 
profile of 'INTP'; people who are on the rdfweb-dev list. FOAF, or 
FOAF+extensions, is adequate for providing RDF descriptions of people 
and their characteristics relevant to such groupings.

The latter notion of 'group' is also useful, but some of the entities it 
picks out may or may not correspond to the more community-oriented
notion of group we see in (i). RDF and FOAF can represent both, and 
people have asked for both, but the details need working through.

This is complex in part for boring technical reasons. There is a
language, W3C OWL (Web Ontology Language) that is good at representing 
the necessary and sufficient conditions for class membership. You could
tell an OWL-aware system something like 'there is a class mygroup1 of 
individuals that are foaf:Persons and who are members of the mailing
list mailto:rdfweb-dev at vapours.rdfweb.org' and the tools should be
capable of answering questions about that 'group'. So much of our work 
will involve making sure what we do fits with the facilities offered by
the (new and complex) OWL language.

Representing these various notions of 'group' is pretty important, and 
something I want to get right. But at moment it is on hold while I 
work on getting a more spec-like FOAF spec out there. Spent a lot of 
last night on that :)

http://rdfweb.org/issues/show_bug.cgi?id=8 is the bugzilla entry for the
'foaf group' problem. If you've use cases etc for what you'd like to be
able to represent, note them here or directly with bugzilla...

> 5.	FOAFster domain?

I had been planning to use this for something in the P2P-ish,
file-sharing for FOAFs space, http://openp2p.com/lpt/a/554 sorta thing. 
At the moment that is complete vapourware as FOAF itself needs deploying 
first, but I'm also worrying lately about using the 'foafster' name at all, since 
it could feed the 'FOAF is just a Friendster knockoff' meme, and I
think that's a rather easy and misleading peg for people to hang us on.
(eg. http://www.wired.com/news/culture/0,1284,59650,00.html this week)

So I guess the answer is that I'd rather keep it in my backpocket for
now, sorry about that. Hopefully you've got some other names bubbling

BTW what sort of schedule are you working to for this project? I'm
trying to learn folks FOAF-related deadlines to plan things out for the
summer and autumn... 



More information about the foaf-dev mailing list