Hello Blue,
"the Internet infrastructure consists of a lot of different technologies
sort of stuck together with scotch tape and chewing gum ~ copper
telephone wires, coaxial cable, and fiber optics, with a bewildering array
of connectors, and, to the extent that I can understand it, a bewildering
array of ways of communicating."
You are not far off from the mark with that description. Let's see if we can
rephrase it a bit, while describing how some of these things actually work.
Today's Internet --at least as we know it in the higher developed regions
of the world-- consists of thousands of autonomous networks which
employ a standard set of protocols which are found in the TCP/IP
"suite."
Those which have not evolved to all IP yet (of which there are many), or
those who have elected to employ proprietary protocols and use their
own architectural designs in their interiors as a means of locking
in customers (let's face it, that's why it's done) must use translation
services at their borders in order to speak with one another, and with
networks which do employ an open form of IP.
For example, IP traffic on a Token Ring LAN may speak with IP clients
on an Ethernet LAN through the use of a router. Likewise, IP clients in a
cable TV network (like RR or ATHM) may speak with other ISPs'
networks in a similar way, provided they use routers and stick to IETF
rules.
Router card interfaces employ any number of different protocol types for
this very reason. They can join together dissimilar network fabrics,
effectively neutralizing the differences between them. Thus, the mish-mash
you referred to never knows the difference.
Stated another way, these disparate networks are able to communicate
with one another on a peer basis through the use of gateways, as
they were once called, but which are now more commonly known as
routers.
Among the various interfaces that can be found on routers today are the
types which also support fiber: Fast Ethernet, GbE, FDDI, SONET, IP
over SONET (called packet over SONET, or POS), and eventually IP
over lambda with a thin shim of either SONET or Ethernet, or none at all.
I'm getting ahead of myself here, but keep the foregoing in mind, for later
reference when I step forward with my own defense. -g-
All of these conditions assume that each of these autonomous networks
abides by an approved list of practices and protocols which are
prescribed by the IETF (Internet Engineering Task Force).
A common notion today is that the Internet has always employed the
Internet Protocol (IP) suite as it is now known, ever since its inception.
This isn't true. During the Seventies and early Eighties IBM protocols
were prevalent, as were X.25 and a collection of switched low speed line
protocols at.. 1.2 kb/s, 2.4 kb/s, eventually dedicated 9.6kb/s and 56kb/s
lines, until the mid Eighties when the first T1s made there way onto the
scene.
IBM was very prominent in early user networks, because most
universities and research organizations were employing mainframes and
dumb terminals back then, prior to the advent of PCs, LANs and client
server architectures. Also used as the defining protocol for these local
nets was DECnet when LANs become fashionable, even WANGnet, and
heterogeneous cable TV types of networks employing frequency division
multiplexing over thick cable coax in large enterprise settings. On the
WAN, IBM was evolving their SNA, DEC their DNA, WANG had their
own plumbing arrangements, and the telcos had their private lines.
X.25 was very dominant in its role as the inter-networking protocol at
first, especially in on line services such as Compuserve, and in
international applications, as it still is today along with Frame Relay to a
large degree. For the vast majority of all present day routes in developed
countries, however, X.25 has been replaced by TCP/IP, or IP.
"When you and ahhaha are talking about "pure" fiber, are you talking
about attaching it to the above-described mish-mash, or do you envision it
being "pure" from one end to the other? If so, wouldn't that be like
starting from scratch, which we don't seem to ever do?"
Actually, the goals which you scribe as being the contrivances of AHhaha
and myself are actually being accomplished right now, not in subscriber
networks (HFC, DSL, Dialup, etc.) at the end points, but in the very
core of the Internet, itself.
In other words, it's not happening at the subscriber network level yet, or
to individual residences. Instead, it's happening at the Internet's core
which is the very center of the 'net (figuratively speaking.. core connotes a
hierarchical reference only, and not <necessarily> some geographic one,
as is often conjured when one hears the term 'core'), and on the
high-capacity edge-to-edge connections now being supported by
very-high capacity optical routers employing dwdm.
It's working its way from the center out to the distribution networks in the
Cable Modem HFC networks and the telco SONET and ATM
networks. It's only in the very last mile where it falls short.
Of course, I've oversimplified much of this, but it's true that all of the high
capacity optical devices which have come onto the scene in the past two
years (such as Juniper, Sycamore, Ciena, Nortel DWDM, LU's
LambdaRouter, Monterey, etc etc) are supporting Internet traffic over the
same forms of "pure" as we've been discussing, only at network's core. A
connection from the edge, or from the subscriber distribution networks
would look no different, since they too would simply need to be attached
at the proper router ports which would be designed for them at that time.
HTH.
BTW, there is no consensus as to exactly what that protocol which would
be used for supporting "pure" at the router port level would be. In fact,
the introduction of a protocol by any electronic means in itself results in
contaminating the transparency - pureness - of fiber, which makes
the entire notion of "pure" an oxymoron, from the start. But that's a topic
for another discussion.
Regards, Frank Coluccio |
