Wed, 13 Jan 1999 11:31:35 +0200
From: "Gur Kimchi" <Gur_Kimchi@vocaltec.com>

The fall of the Stupid Network

Many years ago (or two years ago, depends if you live in Internet Time
or not), a landmark white-paper from AT&T has leaked and circulated the
web.  The basic premise was that for the first time, a large
telecommunication company (e.g.  AT&T) was ?getting it?.  ?It?  of
course is ?The Internet?, and the design principle and architectural
concepts behind it.

Unfortunately, that dream is now beginning to fade, and while the
immediate loss is small, the long term damage is only too frightening.

So what happened?  First a short overview of what are the ?Internet
Design and Architecture principles?, so you know what is being broken.

1.  The Network is dumb, the edges are smart:  In other words, the
network itself does not know what it is carrying.  It has no concept of
?This is an audio call?  Vs.  ?this is a bank transaction?  - it is
stupid, and rightly so - because stupid means fast and cheap.  The edges
on the other hand are where the ?smarts?  is.  In every-day terms, one
?edge?  may be the PC on your desk, and the other is a WWW server
sitting at yahoo.com - the network connecting your PC to yahoo does not
know (or care) what is transmitted, it only cares about getting the
information from side to side.  2.  Push execution to the edges:  in any
system, there are many clients, and much fewer servers.  The more
processing we can do on the clients, the more implicit CPU power we have
at our disposal.  In the old mainframe days, the only way to add
computing power was to buy a bigger mainframe.  In the near future, we
will just add more computers to the network.  This also has an important
advantage - when the mainframe failed, all the terminals stopped
working.  But when a single edge computer fails, only a small segment is
effected, making the system much more robust.  3.  Simple, statefull
protocols:  Most of the protocols used on the internet have an
interesting property:  There is full state at both sides of the
client/server relationship.  This is important if you want your edge of
the communication to make intelligent decisions - if it has state, it
knows what is going on, and can make these decisions correctly.  So
where did these three concepts get misused?  What is the evil force that
is making the network smart and the edges stupid?  - Well, blame it all
on Voice-over-IP - the field that Israel?s VocalTec Communciations
pioneered in 1995.

Voice-over-IP (or VoIP for short) is a wonderful concept - it lets you
place voice calls (and video calls, and other types, but basic voice
service is the problem) over IP networks, using commodity IP equipment
that is much cheaper then the conventional PBXs and Switches of your
telephone company.  This makes the service easy to build, and to the
horror of many telephone companies, many newcomers can build a telephone
network faster, cheaper and with more features then the conventional
Telcos can hope to achieve.

But your mother?s telephone company is not stupid, it can see what is
going on, and because it both has amazing resources, and 99% of the
current customer base, it will try to compete.

So they do - and to compete they have two options, either start anew
with the new ?Pure IP?  telephony, or find a middle ground - one that
will let them use most of the equipment that they have today, and still
lower costs and add services quickly.

An interesting point about Telcos - most of the expenditure for building
a telephone network is not the telephone switches, but rather building
the services and billing ?back-end?  network.  Most telephone companies
spend over 60% of their investments on these parts of the network - if
we could find a way to keep parts of the network useful, and still
support VoIP, well, you will have a winner.

And a few companies (that will remain nameless) did just that.  They
decided to provide a method to keep the signalling and billing system in
place (what the Telcos call the Signalling-System-7 network, or SS7 for
short), but still support VoIP.

And many other companies followed.  The Telco market is so large that
many IP equipment companies accountants practically slobbered at the
idea of getting a foot in the door.  We are talking checks bigger then
the yearly product of many small countries.

But there is a dark side to these advances.  The invention that allow to
mix your grandma?s SS7 network with the new world of VoIP also means
that you will never have state in the edge of the network.  As a
consumer, your telephone can never be smart.  You PC can never decide to
take a voice message instead of ringing when your mother-in-law is
calling.  Granted, your phone company would gladly supply these
services, for a nice fee of course.

So why do so many users have to suffer because companies are terrified
that their market is vanishing?  Because they can.  Because in the age
of the Internet, standards can be developed in the IETF (the Internet
Equivalent of the Legislature), and IETF standards, unlike standards
developed by other bodies like the UN?s ITU-T or the European
Commission?s ETSI, do not need 100% consensus.

Consensus goes like this:  Lets say you have an idea for a standard. 
You write it up in a nice paper, and submit it to a relevant
standardisation body - The ITU-T for example.  The paper will be
presented at the next meeting and read by the members, and if someone
opposes, you will have to compromise.  You will change your idea
slightly, your opposition will change theirs slightly, and in the end,
you will find a middle ground.  And you have to compromise - because if
you don?t, there is no 100% consensus, and an agreement will not be
reached.

100% consensus means that if an 800-pound gorilla company has an idea,
they have to listen to you - and if they want it approved by the
standardisation body, they will have to accommodate your concerns.

But the IETF does not require 100% consensus.  The instead require a
?rough?  consensus, which means that you can be ignored.  A company can
bully it?s way thru and get whatever it wants approved.

In addition, there are valid technical reasons for making your telephone
(and your PC) stupid - this way new services may be added quickly, as
only the server has to be modified.  But of course this also means you
can NEVER have new services on your telephone - the poor thing just has
no idea what is going on, so has nothing to base decisions on.

Flexibility is the key.  The dark forces will like you to believe that
the only way to provide their model (master/slave) is by using their new
proposed standards - but this is plain wrong.  Peer-to-Peer protocols,
if they are well-designed (and in this case - they are) are fully
capable of providing the same functionality.  The Edges do not have to
be killed to make the network intelligent - many solutions available in
the market today provide the same capabilities.

But all is not dark yet.  We failed to show one effect:  That users do
have a choice, and as long as both types of networks will exist, the
users will be able to choose.  And for me that is a simple decision to
make - if two companies want to provide me with telephone service, I
will choose the one the lets me make the decisions at my edge.

For service providers, this is a difficult time.  They will want to use
their current systems - they already spent amazing amount of money to
build them.  They will want the ability to totally control the network
services, but they will be frightened by other service providers that
will allow the customer to choose - making them more appealing.  I do
not envy them.

Do not take this lightly- this is a full scale war.  And the winner will
control your telecommunication services and the ultimate answer - can
your phone decide for itself, or will big brother forever rule?  Will
the Future belong to the Edges or to the Network?

Persona Dramatic:

H.323 and SIP:  Full-State VoIP protocols - the forces of good

MGCP: Master/Slave VoIP protocol - the forces of evil

SS7:  Your mothers ?smart?  network - it drives the world?s telephone
system.  It?s big, it?s expensive, it?s also pretty limited.

ITU-T:  International Telecommunication Union, a UN branch working on
Telecommunication Standards

ETSI:  The European Telecommunications Standardisation Institute

IETF: Internet Engineering Task Force