Interconnected computing

["Dave Farber" <dave () farber net>]

---------- Forwarded message ---------
From: Bob Frankston <Bob19-0501 () bobf frankston com>
Date: Tue, Jan 2, 2018 at 9:56 PM
Subject: Interconnected computing
To: David J. J Farber <Dave () farber net>


Dave, watching the neutrality discussion I think back to your quote about
the design of token rings “Could I use a whole bunch of these minicomputers
together to form a more effective computation environment?" Preserving the
transparency of the computer-computer relationships is key. It’s a form of
the end-to-end argument in which solutions are created at the end points
and not inside a network. The application program doesn’t care whether the
“between” is a particular token ring, Ethernet or, anything else.

Token rings and Ethernets were not services – they were simply shared
facilities. We had transparency between the computers (end points) because
the community (such as a college campus) would pay for the facilities as a
whole. It just didn’t make sense to charge for each connect, especially
since the end points themselves were shared facilities – timeshared
computers.

The remarkable achievement of the internetworking effort was to extend this
transparency across the world. On a LAN the sender could simply retry until
all packets were received. TCP builds on this simple idea, adapting it for
a global network with limited capacity. The interconnections were done
using dedicated wires (or fibers) when available and capacity purchased
from existing telecommunications providers in most cases.

This ability to use existing facilities as a resource is similar to the way
teletypes and later computers employed modems to use a voice network for
data. It was never a good match and in the 1990’s dialup calls tied up
circuits for long periods of time. This created a situation similar to
Mother’s Day when people made many phone calls. There were only so many
lines available. That scarcity was real but an artifact of the
telecommunications algorithm of reserving capacity for circuits (or
channels or pipes) as a way to implement services. Fortunately, dialup
modems gave way to native internet connections before the problem reached a
crisis point. Spectrum allocation is another face of the reserved capacity
technique.

The internetworking approach (best efforts) doesn’t rely on reserving
capacity and uses protocols like TCP for cooperative sharing. I purposely
said, “The internetworking approach” to avoid saying “The Internet” because
the term makes it seem as if the Internet is a physical network or a place
you access. That said …

The Internet has been positioned as a telecommunications service by the
accident of history. In the past computers on campus networks would simply
be connected to the local network (LAN) and that network would be
interconnected to the rest of the world. But users at home would dial up
commercial or campus computers which were connected to the Internet. Modern
infrastructure is designed using Internet protocols as the native
transport. You don’t need make a separate telecommunication connection to
reach the Internet. This means the Internet is the infrastructure. But we
still act as if we need a third party, the ISP, in the middle.

Even when we aren’t yet using IP as the native transport a community such
as an apartment house can bring IP connectivity to the edge buying capacity
from a traditional provider and then sharing it locally as infrastructure.

This creates problems for the traditional telecommunications business
model. One is that they can no longer rely on the revenue from applications
which can now be implemented in apps and over IP. The other is that IP
packets don’t rely on reserved capacity or pipes. The packets can take any
path because they are reassembled at the end points. This means we don’t
need a dumb pipe – just available facilities.

The battle over neutrality is about how pipes are managed and about
preserving the revenue from services. Instead we need to look ahead to a
native Internet-based infrastructure. Without a monthly fee for each
connection we can assume ambient connectivity or “just works” connectivity.
The kind of connectivity we get in connecting our computers directly to
LANs.

I understand that people are concerned about cities acting as
telecommunications providers playing favorites. But they are just
transporting encrypted packets rather than providing services, so they
aren’t in the position to pick favorites. Such approaches are inherently
neutral. I have an FAQ which answers these and other concerns in more
detail.

The infrastructure approach honors the initial vison of creating a way for
computers to join to form a common environment.



Bob Frankston

http://Frankston.com

@BobFrankston



-------------------------------------------
Archives: https://www.listbox.com/member/archive/247/=now
RSS Feed: https://www.listbox.com/member/archive/rss/247/18849915-ae8fa580
Modify Your Subscription: https://www.listbox.com/member/?member_id=18849915&id_secret=18849915-aa268125
Unsubscribe Now: 
https://www.listbox.com/unsubscribe/?member_id=18849915&id_secret=18849915-32545cb4&post_id=20180102221931:E7E54B4C-F034-11E7-9B44-A58408FC42AE
Powered by Listbox: http://www.listbox.com