Re: Skype asks FCC to open up cellular networks

[David Farber <dave () farber net>]

Begin forwarded message:

From: Brett Glass <brett () lariat net>
Date: February 27, 2007 12:47:09 AM JST
To: Brad Templeton <btm () templetons com>, David Farber <dave () farber net>
Subject: Re: [IP] Re: Skype asks FCC to open up cellular networks

Dave:

Some comments on Brad's remarks, for IP, if you will.

At 02:19 AM 2/26/2007, Brad Templeton wrote:

> The problem occurs when users are behind NAT, which happens by
> their own choice, or the choice of their facility, or because their
> ISP does not provide them with enough routable IP addresses (or for
> example, IPv6 which some P2P apps can use though I am unsure if
> Skype does.)

Or for security. NAT makes it much harder to invade users' systems.

Static IP addresses are also expensive, because IANA artificially
inflates their cost. This is especially true for small ISPs.

> In this situation sometimes the P2P applications are forced to
> not route the traffic end-to-end as they would like to.

The could -- as other programs do -- route traffic end to end
without stealing bandwidth. But this is not the case with Skype.
When it finds a system with lots of avaiable bandwidth, it creates
a "supernode" which consumes many resources.

See the following links:

http://computerworld.co.nz/news.nsf/news/ 
7AB67323D6305E49CC2570A1001698C0

http://forum.skype.com/index.php?showtopic=16251

http://gigaom.com/2006/01/10/skype-the-bandwidth-hog/

[SNIP]

> While this is fairly well known, and I think spelled
> out in the Skype EULA,

The user's ISP doesn't sign the Skype EULA. And the user himself
will virtually never read it and would most likely fail to
understand the implications if he or she did.

> For most ISPs, having the relay be on the same ISP is a big
> win for the user and causes minimal trouble for the ISP.

I beg to differ on several points. Firstly, having several Skype
supernodes will cause trouble for ANY ISP, wired or wireless,
and will consumer a substantial amount of expensive resources
for which Skype is not paying. The problem is especially severe
if the supernodes are on the same cable segment (in the case
of a cable operator) but is significant even if they are not.

> Brett runs an unusual ISP which is based on wireless. For him,
> it is not a win.

Wireless ISPs are not unusual. It has been conservatively
estimated that there are now more than 9,000 in the US alone.
What's more, the problems I've cited are not at all specific
to wireless networks.

> I must admit that sometimes Brett seems to act as
> though all network applications should be designed with the
> constraints of a wireless ISP in mind.

In this case, none of the constraints are unique to a wireless
ISP (though wireless ISPs are, in general, being held back in
capacity and coverage by unwise spectrum policy).

> While I don't think it
> is bad to think about those constraints, it is wrong to declare
> applications as evil because they are designed with the
> constraints of wired ISPs.   Wireless ISPs do not have full
> duplex transmission, they can't do upstream and downstream at
> the same instant,

Mr. Templeton may not be aware of this, but cable modems are
half duplex as well.

What's more, full duplex versus half duplex is not a limiting
factor. In many media, in fact, half duplex is more efficient
and lowers equipment costs, giving more "bang for the buck."

Bottom line: the problem with Skype is that it robs bandwidth
without the consent of the ISP or the knowing consent of the
user.

--Brett Glass



-------------------------------------------
Archives: http://v2.listbox.com/member/archive/247/@now
Powered by Listbox: http://www.listbox.com