faq fodder: archiving answers to hugo's questions
danbri at w...
Wed Oct 30 11:13:27 UTC 2002
fwd'd with permission...
ps. sorry i've not replied to a few rdfweb-dev msgs; been a bit busy, will
get to them asap...
---------- Forwarded message ----------
Date: Wed, 30 Oct 2002 05:48:24 -0500 (EST)
From: Dan Brickley <danbri at t...>
To: Hugo Haas <hugo at l...>
Subject: Re: I added Bruce Perens to foaf!
On Wed, 30 Oct 2002, Hugo Haas wrote:
> Thanks for the pointers.
> * Dan Brickley <danbri at w...> [2002-10-30 04:52-0500]
> > > 1. http://larve.net/people/hugo/2002/10/foaf.rdf
> > cool.
> > Opera says: XML parsing failed: mismatched tag (36:85) though.
> > <foaf:schoolHomepage rdf:resource="http://www.ecp.fr/"> isn't closed, at
> > least.
> Yes, I just noticed that. Fixed it.
> I am wondering something: I see that people usually list only a few
> people that they know, and not all of them (e.g. you don't say you
> know me).
> Is that:
> - because you shouldn't list too many people.
List as many as you like...
We don't give a careful definition to foaf:knows because it's impossible.
We originally had foaf:knows, foaf:knowsWell, foaf:friend, but that was
awful, and unworkable. You can't force people to classify others that way
and expect it to work. Instead, we now use foaf:knows pretty liberally,
and concentrate on other more interesting kind of facts: who wrote what
with who, who is in which picture with who, etc. That gets us away from
having to over-classify relationships, creates more useful data, and makes
the anti-AI point that the semantic web isn't about creating super-formal
defintions for relationship types...
> - because people are lazy.
That too :)
It makes another SemWeb architecture point too, about not making 'closed
world' assumptions. Just because some fact isn't in a particular document
(or database), don't assume negation-by-failure, ie. that it is false.
OTher systems (Prolog, SQL) do make such an assumption, typically, but it
doesn't work in the Web, where data is smeared across 1000s of sources.
So on my FOAF t-shirt, it _looks_ like I'm really friends, whereas Mags
and Liz (ex-housemate here) don't appear to know anyone. But that's just
cos the t-shirt only represents part of the picture. You can't assume from
the absense of people being listed on the t-shirt (or their foaf file)
that they don't know them.
> - because you only list people with whom you exchange more than 10
> emails a week.
that'd be a reasonable rule of thumb.
Some of the 6-degrees experiments have used a defintion like 'have shook
hands with', 'is on first name terms with', but all of these are
culturally biased (would they work in Japan? with non-geeks? in 5 years
So in a sense, we're fleeing from such pedantry and concentrating the
tools around stuff that's easier to define: 'does this photo depict
Hugo?'; 'who attended this meeting?', etc. The social stuff is
sorta-implied by these clearer facts, but only in bulk. If I'm in 500 pics
with you, you might infer we know each other. Or that I'm a stalker ;-)
> - completely random.
Not quite random, but not easy to define.
foaf:school is similar. When I created it, I meant 'school' in the english
sense of ages 5-18ish. Americans tend to use it for University, College
etc too. I'm happy it being a bit murky...
> - another reason?
> Basically, should I try to generate a massive file or just try to link
> myself to the Web (just added a pointer to your FOAF data for
Yeah, just add a few folk in, likely those who themselves have FOAF/RDF
The coolest thing about FOAF is something that's in RDF Schema not in FOAF
itself: the rdfs:seeAlso relation, which allows these things to be linked
together and harvested.
Don't even think of asking me why we don't use X-Link ;-)
Oh, slgihtly fun but a bit broken, the world's worst FOAF harvester:
(it's a single perl script. Just run it and watch...)
ps. can I copy your msg and my reply to public www space sometime? must...
More information about the foaf-dev