Date: Mon, 17 Feb 92 17:45:26 -0800
From: karn@UCSD.EDU (Phil Karn)
Subject: Re: When Did the LEC's Start to Die?

I've done a lot of thinking on exactly this point, but with regard to
data rather than voice. The telcos do a pretty good job with local
voice, and it would be hard to compete with them on either price or
quality. As Lauren Weinstein has said, "Your average cable TV company
makes General Telephone look good!"

The real problem is data. Having worked for Bell Labs and Bellcore for
over a decade, I can say that the local telcos still simply haven't
got a clue about data. Either they don't understand the needs of data
users, or they refuse to give them what they want because it's so
different from what they've been selling for over 100 years.

The most successful telco data service is raw point-to-point digital
transmission. Even though it's usually outrageously overpriced, it's
successful for two reasons. First, the telcos have little, if any,
competition. Second, the service is so simple that the telcos have
little opportunity to screw it up. They just have to move the bits
from point A to point B as reliably as possible. Nothing more, nothing
less.

Other service providers, who *do* understand data, can use the telco's
digital leased lines to build the kind of packet-switched networks
that their users really want. The Internet's backbone and regional
networks are the best example, but there are also many private and
corporate packet-switched networks.

But when the telcos themselves get involved in data switching, you get
brain-damaged circuit-switched services like ISDN. Tell them you want
packet switching, and they'll give you X.25. And even these clumsy
services have gone basically nowhere, mainly because of the threat
they pose to the telco's own leased-line business. Why should they
make ISDN universally available at, say, $10 or $20/month and destroy
their own lucrative 56kb DDS markets? It just won't happen as long as
local data transmission remains a de-facto telco monopoly. ISDN has
been "almost here" for over a decade now. If it ever does "arrive", it
will be long after the dialup modem manufacturers finally hit the
Shannon limit of an analog phone line. Nobody will care.

I agree that radio is perhaps the best way to challenge the telcos in
the near term. This is especially true with data because data networks
permit a much wider range of architectures. A voice network is
severely constrained by limits on end-to-end delay of at most a few
hundred milliseconds. This precludes many interesting possibilities,
such as multi-station relaying. But other metrics, such as throughput
and cost, are usually much more important than delay in a data
network.

You could easily conceive of a metropolitan area packet radio network
in which each station not only generates and sinks its own traffic,
but also relays the transmissions of its neighbors. When communicating
across town, your packets would be relayed through intermediate
stations instead of being sent directly to the destination. Although
this increases delay, this approach minimizes transmitter powers and
consequently the interference to stations at other locations in the
network that are trying to send their traffic at the same time. The
result is a much greater overall network capacity than a network that
relies on direct point-to-point communications between user terminals
and relatively distant, central "hubs", as in traditional cellular
telephony.  As the station density in a given area increases, the
average distance between neighbors decreases and so does the average
power of a transmission. This effectively increases the carrying
capacity of the network to help accomodate the additional users.

The network could also be augmented by a few well-placed point-to-
point transmission links (microwave or fiber) spanning relatively
large distances within the network. This would offload some of the
traffic that would otherwise flow across the radio network and again
increase its carrying capacity. Because these links would be shared by
all the users of the network, the per-user cost would be small even if
overpriced telco facilities were used.

Even better, the system is highly decentralized, making it more robust
against the failure of critical nodes and much more difficult for one
powerful entity (like a telco) to exercise control. Nodes providing
services primarily for the benefit of others (such as the cross-network 
"wormholes" and gateways to long-haul networks) might carry only the
traffic of users who agree to pay for their use. These services
wouldn't be monopolies -- if someone charged too much, anyone else
would be free to provide a competing service at a lower price.  But I
would hope that most of the network would operate as USENET does now,
with each node providing free relay services in exchange for being
able to use the rest of the network to carry its own traffic.

This is a radical concept, but I think it is entirely realizable. Much
of the basic technology has already been pioneered primarily by the
DARPA SURAN (SURvivable RAdio Network) project, although SURAN has yet
to prove that the equipment can be made at commercially viable (as
opposed to militarily viable) prices. Getting the necessary radio
spectrum is another challenge. PCS at 1.8 GHz is one possibility, but
even now Part 15 of the FCC rules allow unlicensed one watt spread
spectrum transmitters on certain UHF frequency bands. Equipment is
already available to operate under these rules, but in my opinion it
is all either junk or overpriced. The market is still wide open to
anyone who can build a truly high performance, low cost Part 15 spread
spectrum transceiver.

I confess to some doubts in advocating radio bypass of the telephone
companies, as radio spectrum is a very precious resource. It is
usually preferable to use copper or fiber for fixed applications,
reserving radio spectrum for mobile communications. But a local packet
radio service such as the one I have described will have served its
purpose even if, God forbid, it is eventually overtaken by a
well-designed wire or fiber packet switching service provided by the
telcos. At least it will have finally woken up the telcos to the
urgent need for low cost, high performance local packet switched data
services.

But then again, one might as well have tried to get Western Union into
the telephone business.


Phil

From: karn@chicago.qualcomm.com (Phil Karn)
Subject: Re: Telco Data Services
Organization: Very little
Date: Sat, 22 Feb 1992 20:57:44 GMT

In article <telecom12.159.5@eecs.nwu.edu> kentrox!bud@uunet.uu.net writes:

> I'm not at all sure of the telco's dedication to the concept, but have
> you looked at SMDS?

Yes, I am quite familiar with SMDS. A couple of years before I left
Bellcore, I participated in a rather in-depth internal review of SMDS.

In theory, SMDS is quite nice.  The details are a little too
contaminated by IEEE 802.6 and ATM brain-damage -- there's a lot of
gratuitous complexity that will make the system much harder to
implement than it need be -- but unlike ISDN, the service model is
fundamentally sound. (Aside: why can't people learn that when
standards committees screw up, as they so often do, the proper thing
to do is to IGNORE them?  If you slavishly accept a standard, no
matter how broken, just because it came from an august body like
CCITT, ISO or TIA, then you're just inviting them to come back and do
it to you again and again. Somebody has to teach the standards
committees that their work will have to earn the right to be called a
"standard".)

Anyway, the real question, as you allude, is the telco's dedication to
the concept.  Sure, there's been plenty of lip service. But as with
ISDN, I've seen little to give me hope that this service will really
become ubiquitous and affordable. Also, remember that SMDS is being
targeted for large multi-location companies, ones who can already
afford T-1 or faster leased lines between their facilities. I can
guarantee you that you won't be able to afford SMDS service to your
house.

The root problem, of course, is that the telcos have virtually no
competition to spur them on. They have little motivation to provide
SMDS any time soon, especially since the main "competition" is from
(and to) their own leased-line services. (The commercial components of
the Internet are too small to be meaningful to the telcos, and in any
event they generally provide inter-lata service, which is barred to
the telcos under the terms of the MFJ.)

Don't misunderstand me, I would *love* to be proved wrong on all this.
In the rather unlikely event that the telcos were to see my note and
be so ashamed by it that they were to immediately tariff universal
ISDN and SMDS at reasonable rates, I would be absolutely delighted.
I'd sign up tomorrow.

But I'm not holding my breath, and that's why I'm advocating a packet
radio data service that bypasses the telcos entirely. Because even if
it lasts only until the telcos are spurred to action, it will have
been well worth it. And if the telcos ignore it, well, then at least
we'd have a reasonable data network to use.


Phil

PS: Needless to say, these are strictly my own personal opinions.