[rdfweb-dev] RE: [Geowanking] Scraping geolocated friends from Friendster

Ben Discoe ben at vterrain.org
Wed Aug 6 21:57:07 UTC 2003

> From: Dan Brickley [mailto:danbri at w3.org] 
> Cool, thanks. The graphviz layout looks interesting. I'd be 
> running from Linux, so I guess I'd need to request VTP and
> build that first.

Mmmm.. yeah.  The good news is that the vtdata library is the smallest
part of the VTP and it's the simplest to build.

> But just having the code to look at is definitely handy

Yup, it ain't brain science.  It took 2 days to write just because i had
to learn a lot about libwww and cookies along the way..

> maybe someone from the FOAF list will be into running
> this on windows, transliterating into Perl etc.,
> or make a Linux binary available...

The only significant external dependency is GetURL(), so a port to Perl
or python or whatnot would just need to drop in the local equivalent.

> One more request: could you make the .dot GraphViz file 
> available that generated 
> http://washedashore.com/people/friendster/neato1.png

The only difference from the default neato behavior is changing the
shape of the nodes:
  node [shape=box,height=.1,width=.2,fontname=Arial,fontsize=8];
Ah yes, i also made the tier1 friendship links longer [len=2] than the
rest.  The rest of the layout is just what neato does by default.

For the record, i'm not happy with neato, the "all points repel each
other" algorithm it uses leads to dumb, unnecessary overlap (e.g. Daisy
and Christopher in the top right of that example above.)  Seems to me
that in a force-based simulated annealing approach, the force should
drop to 0 after a small distance, so that distant piles of nodes don't
have a bogus cumulative effect.  But i don't have the time to get into


More information about the foaf-dev mailing list