A problem for "Web as distributed programming paradigm"

Jeff Bone jbone@jump.net
Fri, 24 Aug 2001 11:40:52 -0500

Clay Shirky wrote:

> So now I have to figure out which of those three possibilities it is,
> but I'm leaning right now to making the argument that 'encapsulated
> complexity plus restricted channels of information transfer' is an
> actual deep connection between any multi-agent system, whether
> biological or computational (or juridical, for that matter, as with
> company to company communication.)

One other thing to consider that plays into this analogy;  intercell
communication in the body is accomplished by a number of parallel systems (such
as the ion transfer that you mention mediated by the presence of
neurotransmitters, the cytokine signalling system used by the immune system,
etc.) but there's a fair amount of commonality.  In many of these systems, the
signalling "fabric" is described by a gradient of available chemical agents.

This is significant;  it wouldn't be totally wrong to say that many of these
signalling systems use the ambient levels of appropriate substances as an
external data store;  they look (if you squint some) like the shared-space
coordination systems and languages we've been talking about.  That's key:  the
individual communicating agents (cells) are not tightly coupled and do not
communicate point-to-point or type-to-type, but rather exchange data by
emitting and absorbing appropriately "formatted" chemical information, which
otherwise just hangs around until it's removed.

Pushing the analogy further, the space of things named by URI and accessible
via HTTP is this shared environment which mediates the communication, and the
particular model of a given resource representation for the various HTTP
methods is equivalent to the particular model of a chemical signal.

The analogy breaks down a bit in that while cell communication is entirely
associative / generative, the Web still has a notion of addresses.
Communication endpoints are still described, even in REST, with URI, which when
interpreted as URLs leads to tighter coupling than the biological model.  (In a
more generic interpretation, say in an HTTP with an application-level
resolution / routing mechanism, this coupling is loosened but the associative
quality of access is still constrained to "this thing with this identity.")  In
that respect, cell biology is much more like Linda than the Web.

Random thoughts,