[E-Lang] SocialModels4P2PNetworks

[puff@guild.net]

Mark Miller writes:
> (And yes, egroups really does truncate the author's (Wesley's) email address 
> at the top and bottom of the message, presumably in order to inhibit you 
> from contacting the author through a non-Yahoo-based communications medium.  
> WebRing interactions are similarly mutilated.)

     Much as I dislike egroups (several of my favorite mailing lists
moved to egroups, which was then bought by yahoogroups, who are, if
anything, doing an even worse job), this is not their fault.  It's a
fairly common anti-spam measure to protect list members email
addresses from harvesting (for example, I think mailman, one of the
more popular mailing list managers, provides it), and I think it's
even configurable per-user and/or per mailing list at egroups.  

     Personally, while I have plenty of direct experience with the
deluge of spam that can result from participation in public forums,
I'm not sure I really like this approach to dealing with it.  I'm
still trying to think of a better approach, but...

Wesley Felter wrote:
>In systems based on global names, once someone knows your name you can't 
>really prevent them from contacting you. Spam is an example of this problem: 

     The gotcha, I think, lies in the idea that you can only have
either extreme, either everybody can send you mail or only people
you've allowed can send you mail.  Even the latter, even if based on
capabilities, has weaknesses.  If you can only be contacted by people
you know, how do you ever get to know new people?  

     I guess this falls into the realm of "What do we use capabilities
to build?"  

     For years I've been pondering the implications of social
software; email, mailing lists, newsgroups, chat systems, muds, wiki
wiki webs, etc.  Last year I spent a lot of time roughing out some
ideas for a large key-based trust system with inclusive rather than
exclusive filtering.  Imagine if you could set up a system where users
could hand out revokable digital keys to other users, sort of like
signing a PGP key ring, only with some topical context for each key.
Combine this with inclusive filtering - you only let through people
with keys you trust, although the system would need some complex
filtering tools.

     But how to avoid too much insularity and stagnation?  Two
concepts come to mind.

     First, add additional dimensions of interaction.  For example the
idea of social spaces that might overrule or disallow certain
filtering configurations, creating defacto public spaces where people
would be exposed to interaction with others who would not normally
come in contact with them.

     Second, push the level of granularity of the keys down to a very
fine level.  Generally the idea here is to discourage a "litmus test"
approach to filtering, by making it extremely easy to use more
discrete approaches.  I'm still pondering the details of how to do
this - simply making a very usable interface is a start, but not
nearly enough on its own.  There is a natural human tendency to
simplify, which leads to oversimplifying, particularly in situations
like this.  I'm still trying to find a natural human tendency that
I can empower to counteract this tendency.

     So I collected notes and massaged them for about half a year or
more.  Then an accident torched the file full of notes, which quite
annoyed me for a while.  Now I'm revisiting the same ideas, only
instead of a monolithic system, I'm wondering if such a system can be
made in a distributed form.

     This must be where the peer-to-peer stuff comes in.  However,
I've always been extremely pragmatic - while I love to wonder and
speculate about the future of systems and the concepts, I'd like to
focus on something that can be feasibly built in the not-so-distant
future, and part of that means building something that fits into the
way people use systems today.  

     I'm considering taking a first cut at this as a web-based mailing
list archive with collaborative filtering - users get accounts to
access the archives and can then annotate messages with keywords and
can filter messages based on their own keywords and others.  This will
allow me to play around with the ideas and see what people come up
with, but in the end it's monolithic in the extreme.  What's the next
generation? 

     One approach is to abstract it down to the essentials and then
bring it back to everyday use.  In its purest form, what we're talking
about is communication and identity.  Email addresses are convenient
expressions of identity, used for communication.  

     What if every mailing list and every newsgroup or BBS gave you
your own email address, particular to that forum?  Not just that
specific provider or type of forum, but that particular topic?  What
if this happened so pervasively and effortlessly that it became built
into the software we use constantly for sending and receiving email
and reading newsgroups and posting to them?  

     For this to happen, two things have to change.  First, our sense
of identity and reputation have to be separated from our sense of "how
to contact somebody".  Second, our idea of what an email address and
how it's used has to be pushed down to a finer-grained, more
underlying level.  

     The net result of both of these is to make the transposition from
"some member of this mailing list said XYZZY" to "some objective
person with a given reputation said XYZZY on this mailing list" to
"I'm attempting to contact this person through my involvement in this
mailing list" to "my system is sending bytes to your system that your
system will empower with some amount of your attention."

Steven J. Owens
puff@darksleep.com