[Chapter-delegates] On the Bandwidth congestion and the inevitable internet crisis.

[Marcin Cieslak]

Sivasubramanian Muthusamy wrote:
> Dear Chris Grundmann,

Inspired by the NYT article I have decided to do some research on my
own. There is a fascinating discussion going on in the Transport Area
Working Group of the IETF here:

http://thread.gmane.org/gmane.ietf.tsvwg/5184/focus=5199

> 
> The present TCP control works like a free for all and laneless bridge
> for automobile traffic over a wild river. The TCP control dumps a few
> automobiles onto the river when there are too many cars side by side.
> Partial roadblocks on either side of the bridge go up if there are too
> many cars seen thrown into the river. ( The analogy is on the
> fundamentals of the problem - it is not exactly relevant in various
> details. It is not to be misunderstood as advocacy for "control" of
> the Internet )

Motorway analogy is not always working with the Internet. First, it is
the endpoint that has most control of the Internet. Imagine sitting at
home and sending thousands of cars from your command center in the kitchen.

> 
> 3. Most arguments on net neutrality w.r.t. to the P2P traffic
> restrictions by ISPs are  misguided. P2P proponents are unfair in
> claiming that the ISPs are unfair, because it is P2P that is
> technically unfair. P2P came into existence with an exploit of the
> loopholes. The TCP protocol with Van Jacobson's patch did not
> anticipate multiple streams with multiple persistence so there were no
> laws or even conventions on this aspect. P2P was invented in the
> absence of such conventions on the number of streams and the intensity
> of persistence. Technically speaking is founded on a trick.

Nobody told anywhere that the world has to be fair.

Mr Ou of ZDNet writes [1]:

"I asked Mr. Briscoe how it was accepted by the IETF and he explained to
me that they had a straw poll the day after that presentation who would
still define fairness the TCP (Jacobson algorithm) way."

Compare this with response to the same on the IETF list:

``There's a third option not presented that might fit more closely:

"TCP's behavior defines a kind of fairness; staying inside the bounds of
that fairness is what we currently defend, largely because it's deployed
and appears to currently be reasonably sufficient"''[2]

TSVWG members doubt whether "fairness" or "balance" was a design
principle of the Internet, even when talking about van Jacobson's
congestion avoidance algorithm. What's known is that it has proven to be
difficult to implement any kind of Internet control by other means then
at the endpoint - the user himself. For example see history of various
QoS schemes including the RSVP protocol.

Chris Gundemann wrote:
> If so, I would ask what your opinions are on the suggestion in this article
> that this is an engineering problem which has been politicized?

Mr Briscoe proposals, although technically doable, tend to change a very
fundamental principle of the Internet defined as a dumb network of
the intelligent hosts. As someone on the TSVWG list pointed out, the
question is not whether we can do it - the question is whether we
*really want* to do it. And this is a very good question for us to
discuss here.

Recently I tend to believe that the congestion problem publicized
recently comes out of marketing strategy of the broadband ISPs. My
German ISP started offering 16 and 30Mbps downstream packages. Being
myself a heavy Internet user, the only reason to upgrade from 2Mbps to
6Mbps package was increased upstream, so I really begin to wonder where
this push for higher packages for residential customers comes from?

Broadband ISPs are tangled in a kind of race of arms of competition.
It works while your margins are still acceptable. If it doesn't work
anymore, you may try other ways of fixing the problem, like screaming
for more money for Internet infrastructure or try to keep your costs low
by fighting the "abusers".

Having worked few years with a product marketing of a broadband ISP
I hardly find justification for the alarmistic tone of many
publications. There is a question whether assumptions that many ISP are
built on still hold today. The Internet is blooming in many unexpected
ways and the business model from few years ago (typical timeframe for
the investors entering such a business) may no longer hold.

In the city of Warsaw there are two competing cable operators. Their
simplified strategy: one of them offers promotions like "1 year free" to
attract new customers, the second one increases bandwidth available
while keeping prices frozen. I bought the cheapest package there few
years ago and now my package got few times as much bandwidth as before.

Financial analysts praised the strategy of the latter provider, saying
that their strategy does not directly influence their revenue stream, as
opposed to the former one. Looks like the cost side of providing more
bandwidth is not on their radar yet.

There are also other issues, like lack of transparency. I could
recommend further reading from the NANOG mailing list, for example here
[3].

I am skeptical to the whole "Internet governance" crowd. I see ISOC
primary role to be involved there to make sure that no harm will be
done. Thanks to our public policy team for doing this.
Most of the business executives I talk to are not really interested in
those issues. One of them (head of a major ISP) expressed it this way:

"They started to make a big issue out of this who holds root DNS
servers. We are also running a copy, and so what?"

I also cannot believe that something that Jon Postel did for years
almost alone requires a multi-million dollar organization now.

A real change will be forced by economics - whether P2P will drive ISPs
out of business still remains to be seen.

[1] http://blogs.zdnet.com/Ou/?p=1078&page=3
[2] http://thread.gmane.org/gmane.ietf.tsvwg/5197/focus=5214
[3] http://thread.gmane.org/gmane.org.operators.nanog/52658

-- 
              << Marcin Cieslak // saper at saper.info >>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 4382 bytes
Desc: S/MIME Cryptographic Signature
URL: <https://elists.isoc.org/mailman/private/chapter-delegates/attachments/20080329/d24f7b1a/attachment.bin>