On Tue, Sep 24, 2002 at 01:59:35AM +0100, Paul Makepeace said:
> Probably worth putting in an expiry thing - if you ("one")'ve lived in
> a place, that'll never change but you could easily break up with your
> current squeeze and invalidate a link.

I've been meaning to put in something that allows you to put a time span
on a relationship and allows you to expire one later if you'd put it in
as 09/00->present or something. 

The neat thing about this would be that you could generate a new graph
snapshot for every month since the first date in the system and watch as
new sub nodes and clusters spring up and little tendrels (sp?) creep out
to each other.

I'm also wondering if I should make it so that you have to log in and
can only add a link between your entity (possibly making it so that non
people entities are created by someone who becomes responsible for it)
and someone else who is emailed when a link is added to to them. 

 
> This gets into the whole vast topic of trust and reputation metrics. How
> well do you know someone? Are links bidirectional or should it be
> consensual? What meaning to assign to an unreciprocated link?

I did think about having something like that but just figured that was
too complicated, would put people off adding new links and was basically
unnecessary, for my purposes anyway. I just want to see the nebulous
links that there are out there and see if there are any patterns. The
problem with the patterns, as far as I can tell is that their emergence
is subjective to the algorithm that graph drawing program uses. Graphviz
seems to favour trying to stick everything in a flat line and then
looping everything round the outside which isn't quite what I had in
mind. I kind of want clustering.

 
> http://paulm.com/friends/ecademy/ was a frenzied hack mapping a (mostly)
> London business network a while back. Lots of resources at the end.
> Definitely would be open to chatting offline if people get bored here...

Oooh, shiny. I actually did a prototype of what I was trying to do about 5 
or 6 years ago but it crashed a commercial graphing program even then. I've
been meaning to finish the code off for ages. I'm going to read through
your list o' resources and I'd love to talk about this, I'm happy yo take
it off list if people think it might get a bit dolphin noisey.

On Mon, Sep 23, 2002 at 09:05:49PM +0000, Earle Martin said:
> /me wonders if the kind of information people put in their FOAF
> files[0] could come in handy (works for, etc.) or if perhaps I
> shouldn't be thinking about this after beer at 2am.

This is kind of what sparked the conversation last night. This and FOAF,
I think, are trying to solve similarish problems but they're not *that*
overlapping - FOAF is kind of a resource indicator and trust network
thingy wheras what I'm trying to do is graphically map the networks and
clusters that people have built up over the years - nothing more,
nothing less. A FOAF export thing would probably be a better idea than a
FOAF import.


On Tue, Sep 24, 2002 at 07:53:29AM +0100, Roger Burton West said:
> I think you should probably have both. I know how to get in touch with
> everyone who posts to this list; but I don't know, say, Shevek, from
> anything beyond his list posts, whereas I know Nick Cleaton because I
> used to work with him... and the ex-Torrington people I keep up with
> are no longer "ex-Torrington" in my mind but just "people I know".

I could have intensities or strengths on the arcs but, again, this is
complicated and also a bit subjective. 

> I suspect I'd have a description field on the link, which might be
> "common workplace" or "shared house"...

What, you mean, kind of like the 'worked' and 'lived' link types and the
ability to add new link types? :)

Simon




Reply via email to