Thread, this is a rather long and sometimes tedious reply if you are
not interested in matters having to do with the semantics of network
architecture.
This is Part 1 to a two-part reply.
----------
WEBIII,
I would not presume to speak on matters of the "Telecosm" as George
Gilder can and hopefully will. Allow me instead to reply with some general
observations, apart from those of the Telecosm specifically, while you await
a more qualified reply in that regard.
-----------
Your message directly touches upon a subject that I've been laboring with
for quite some time. Most of this post is in direct response to the following
rather astute observations and questions you wrote in the uplink post:
"...at one extreme we have a model where all of the intelligence is built into
the network, and the peripheral devices consist of little more than a tin can
and a string, and at the other end we have the Gilder model, where the
network is dumb as a stone, and all the intelligence has migrated out to the
most extreme periphery... Might the most efficient system be some hybrid
of the two models?"
I would agree with your observation that in fact two different models exist
right now, and probably will continue into the future, but in all likelihood not
for the same reasons that George and perhaps others would agree, although
I don't know that for sure. Maybe George will comment on this later.
We call the two different models you referred to, in the more popular sense
when this level of taxonomy comes into play, as Reference
Models, or RMs.
In this post I'll refer to the ISO/OSI Reference Model, or the OSI-RM.
since it is the OSI RM that most closely governs how telecommunications
systems are designed and viewed.
[[As a note to the uninitiated, the ISO/OSI-RM stands for the Organization
for International Standards (ISO) / Open Systems Interconnection (OSI)
Reference Model (ISO/OSI-RM), or simply, the OSI-RM.
The 'real' OSI-RM consists of a single stack consisting of multiple layers,
Layers 1 through 7. Each layer is said to have sublayers and exists in
various "planes." Each of the upper-lower layer junctions take place through
a process known as the "convergence."
The OSI-RM starts with a physical element at Layer 1 (e.g., wire, fiber,
coax, free space, etc.), and works its way upward to the application layer,
which is at L-7.]]
An RM will define the relationships of processes within a communications
system, while stipulating the relationships of each process to one another.
It's an architectural framework made up of a number of points of
reference, in effect.
An RM can be uniquely viewed depending on the context of a given
individual's station in the universe. Taken a step further, I can see how it
can be regarded as being keyed to a given perspective, or vantage point,
such as that of an end user, or that of a service provider. In any event, the
functions of a RM seem to be able to exist in different (sometimes
conflicting) contexts, when viewed within different planes or dimensions.
Reference models are useful in their utility as mechanisms which allow us
to cope with highly complex models by providing an almost palpable level of
substance with which to work, even if they are largely comprised by only
mental imagery and concepts. Eventually, those concepts get transformed
into tangible devices, since the RM contains the functional relationship
principles by which those devices are designed and manufactured, and
ultimately implemented in live systems.
An RM allows both a starting and finishing point which together allow us to
rationalize how to put complex systems together, and analyze how they
function together.
The network transport and routing related layers within this stack comprise
roughly Layers 2 through Layer 4 (effectively, between 2 and 4), with
increasing degrees of fuzziness below and above Layers 2 and 4,
respectively.
ATM and Frame Relay are said to take place at Layer 2, and IP routing is
said to take place at Layer 3. The fiber-optic layer of communications at
the line code level (where light is launched onto the strand) is viewed as
Layer 1.
If you are an end user, then you normally view the carrier's optical
transport services as a Layer 1 set of functions. Normally, that is.
But if you are a carrier or other form of service provider (SP), then things
change with respect to your view of the universe, because SPs do not
maintain the same perspectives as end users. [ ...yes, I know.]
From their perspective, and depending on the technologies they are using,
SPs too will have various layers of the stack in play at any single point in
time, even though you are viewing their overall presence as a Layer 1
physical network element. This can get tricky, as I hope to first describe
and then delineate, below.
The issues you've raised immediately called to my mind not only some of
the precepts suggested in Gilder's Telecosm, but also the principles that
were enumerated more recently in David Isenberg's "stupid network"
writings at: isen.com
Isenberg talks about the shifts which are taking place right now as we
slowly depart from the public switched telephone network's (PSTN's)
highly centralized intelligence model, to one where the intelligence resides at
the network's edge and end points. More recently, there was another
similar work which I posted here about a month ago ("Netheads versus
Bellheads") which was written for the Canadian Government by
T.M.Denton, François Ménard and again, David Isenberg. If you care to
read up on this work, it can be accessed at:
tmdenton.com
-----------
In my opinion, your reference to "two [reference] models" aligns very
nicely with the realities which are now appearing in the market place as the
SPs begin deployment of their next generation of optical/photonic networks
[SONET/SDH being the first ones, arguably], and as demonstrated by a
number of recently published service offerings and product releases by their
respective vendors.
Whereas the traditional view has been that fiber-optic based media was
comprised of a purely passive characteristic (i.e., dumbness) at the
OSI-RM Layer 1, it has become clearer that an entirely new model of
photonic activity is emerging with its own, self-contained multi-layer
architectures, which are suitable for frame working themselves.
Previously (and admittedly, still), these "multiple-layers" of activity within
the emerging optical networking spaces were and still are considered
merely sublayers of the original OSI-RM's Layer 1.
But more recently, these sublayer network dynamics which we are still
calling sublayers have, in my opinion, outgrown such restrictive and
subordinated characterizations as being sublayers of the OSI-RM's Layer 1,
and have caused me to re-focus on some of their implications, since they
begin to create separate models unto themselves (as the remainder of your
post also suggested).
This second order of RM activity at the SP level has an effect not unlike
that of a 30 frames per second (fps) series of still images being
sequentially flashed onto a screen, which creates an illusion that a seamless
moving image is present. To the end user this means that the SPs fiber
based network is transparent, allowing for near real time delivery of end
user content, despite the fact that their information may be taking an
unknown number of routes within a three second period in time supported by
lightning fast flow management, routing and switching in the SP's photonic
"sublayers."
So too, and in this manner, does the second order of SP RM activity
represent a certain level of spoofing to the end user and their applications,
in other words, as I hope to explain in a moment.
While such a view as I am suggesting here will not REALLY reveal
anything new (i.e., these same observations could be equally stated about
the SONET/SDH or earlier T1, T3 deployments of the past), I think that it
is necessary to expand on the optical paradigm a bit, in order to dispel any
growing notions that the optical layer will be immune to such a rules-based
framework as those of the past were, or that it will result in anything
different, when the service providers and their vendors are done sculpting
it. Save, of course, for its ability to scale to much greater levels of
throughput and payload deliveries.
In their purest forms, silica fibers contain no intelligence, while they are at
the same time viewed as having maximum (some say near infinite) capacity
for information carrying, limited only by I/Os. These qualities make them
ideal as "the" passive medium of choice, which Gilder has fashioned into a
form of mental imagery aid known as "the fiber sphere."
"Fiber sphere" is a term which [I believe] George coined close to a decade
ago, and he has made allusions to it many times since then both directly and
indirectly. Such a concept of having a transparent medium aligns very
nicely, although, somewhat mysteriously, with the OSI-RM's Layer 1, and
the dumb core concepts which were later spelled out by Eisenberg and
others.
I don't dispute the underlying theoretical soundness or the principles of the
fiber sphere. Indeed, I have on many occasions applauded George for
conceiving and writing about it in the manner in which he has. At the same
time I think it's more than simply noteworthy - I think that it is an imperative
- to discuss how the SPs have actually responded to the new potentials
afforded by optical developments, in ways which have become clear, by the
nature of the optical platforms that they and their vendors have chosen to
deploy and produce These lead me to conclude that attaining such a true
"fiber spherical" condition is a long way off, as George himself has also
pointed out today in response to a question regarding a cable modem
vendor's (TERN's) prospects.
Granted, we're still in an early stages as far as next gen optical goes, but
some directions have already been firmly established that will, like
everything else, leave a legacy of behavioral and investment realities which
will not only be difficult to get away from, but will self perpetuate well into
the future.
In order to scale these transparency qualities while continuing to maintain
some semblance of "any-to-any connectivity," however, we find that there
are economic realities [supported by the currently understood laws of
physics] which have already been broached and negotiated by vendors and
SPs, alike. They've concluded that in order to give the appearance of
achieving these ends [transparency and unbounded capacity, for the
moment] they can get away with it at 30 fps in an affordable manner by
introducing intelligence, sometimes vast amounts of it, within Layer 1 of
OSI-RM, itself.
Thus we introduce another stack of functional layers, or activities, at the
optical service provider stratum as yet another multi-layer model within the
original end user's perceived Layer 1 of the first OSI-RM order.
Where the first order (RM-1) addressed end user considerations and their
perspectives, the second order ( RM-2) addresses those of the service
providers, and their perspectives.
END PART 1. PART 2 Immediately Follows. |
