After recent rumination on the STARRYNIGHT project
(http://www.rhizome.org/starrynight), full fledged discussion got
underway with stickman's interesting critique:

STARRYNIGHT is, i agree, both contextually and aesthetically a
commendable project. It is also very problematic in the following way:

It is being created by a group dynamic of art-related discussion. A
discussion, that has been delineated by the creative authors of the
project. This is their right. Yet, such a decision (however justifiable)
places the subject matter "art" within a hermetic envelope which–in my
opinion–neglects much of the discourse within the last 20 years,
including activist and context-oriented site-active art.

The restriction of content seems to me to be a decision that is 1)
unnecessary and 2) dull.

Unneccesary because data is the basis of the net. Additional text-based
content would neither add particularly large amounts of data, nor
detract from the network-map of art-relevant commentary that its authors
wish to highlight. Dull because of the same reasons, and the feeling
that the authors don't really trust their own creation. I would find it
much more stimulating to allow a true constellation to arise, trusting
editorial decisions largely to your very admirable java script.

This is, of course, a personal preference - one that leads me to RAW,
not to an edited version. So i am disappointed 1) to find my decision to
recieve RAW unrespected and 2) to see STARRYNIGHT be created as an
edited site map.

The nature of the net is decidedly anti-hermetic. Recent decisions at
RHIZOME might be interpreted as a step towards the staking out of a turf
called "art." A step towards something we all know. A security blanket.

What responsibilities do the RHIZOME triumverate draw upon themselves
when creating an art work based on community involvement?

Colin Keefe replied:

STARRYNIGHT isn't really intended to be parallel to what happens in RAW
anyway (or at least I imagine not), for the following reasons.

1. It's the cumulative result of searches on an edited database, not on
a RAW mailing list archive.

2. The interests of the searchers and the interests of the content
providers (us) don't necessarily coincide.

3. Topics of lesser interest tend to disappear, as one's attention is
drawn to the more bright, more visible subject matter. This is
qualitatively different from what happens on RAW, where all topics are
given the same priority by the listserv.

I myself would find a "STARRYNIGHT" interface to a raw mailing list
archive, especially RHIZOME RAW, compelling–a cool tool to unexpurgated
content.

I wouldn't call STARRYNIGHT hermetic. But if there's disconnect between
the interface and what happens on RAW, it's because there's already
disconnect between RAW and what gets edited and archived.

Responding to stickman's desire for more interesting, self-organizing
STARRYNIGHT constellations, Martin Wattenberg wrote:

as one of the STARRYNIGHT programmers, i would love to see the applet
create its own spontaneous constellations and groupings… indeed, my
hard drive contains several failed attempts to do just that. alas! it is
a tough task, turning words into a map. but we haven't yet given up
thinking about it… i would be fascinated to hear any thoughts on how
to create a self-organizing visual message board.

some interesting previous work:

http://www.media.mit.edu/~kkarahal/loom/
http://graphics.lcs.mit.edu/~becca/webfan/
(these portray social interaction well, less strong on content)
http://www.newsmaps.com/
(shows content well, less strong on relationships between items)

Frederic Madre replied, offering advise on how to reorganize the backend
database:

i'd suggest to store the information into a relational database like
mysql. the design of the database would include a table for "keywords"
with just that in it or maybe number of texts referenced by it, a table
for the "texts" with url and author maybe and page views (probably
populated by an add-on feature developed over the log files of your
server that would update the database), a table that associates the two
"key-txt" with one keyword and one text. Then the application would only
have to read all the "texts" table and allocate a magnitude of
starriness according to the page views and randomly disperse the stars
on a grid while keeping this in a user-related table "usergrid"
including grid coordinate and text id. each star being related to a
text, when you'd click on it the app would read the "key-txt" table and
dynamically display the keywords found. when a keyword is clicked then
the app would look up this same table and look in the "usergrid" to see
1) which are the related texts using this keyword 2) where the hell it
is on this user's grid. then the app would draw the lines between the
grid points.

well, this is _one_ way… it's a very rough and simple DB design.

Colin Keefe:

Frederic's description is actually pretty straightforward (and a very
common solution to the problem, probably–organizing data is not a
terribly new profession).

While his model does describe a "self-organizing visual message board"
it doesn't provide a method for the messages to be self-organizing in a
*meaningful* way. Meaning that the spatial orientation of the data
serves no purpose in this model (same with the existing STARRYNIGHT, to
my knowledge). Randomly plotting points on a grid isn't as meaningful
as, say, visually clustering by content.

There are a couple ways to add more dimensions to this n-dimensional
representation (and still keep within the STARRYNIGHT metaphor):

1. Cluster like data in galaxies.
2. Allow zooming to home in on increasingly similar (and therefore
spatially more clustered) data.
3. Mark data content by red-shift.

Oh yeah, and why "STARRYNIGHT" anyway? Why not "Creepy RHIZOME?" Why
can't these be rhizomatic root structures? With root nodes getting
larger and more tumor-like with every page access. Maybe they even throb
or something, and new topics send off new tuber-like shoots. Isn't that
a more appropriate metaphor?