[ NNSquad ] Re: New P2P Privacy System from Univ. of Washington

[Wes Felter <wesley@felter.org>]

Richard Bennett wrote:

> Jacobson raises one of the interesting challenges for neutralists, to 
> wit: if you insist that protocols remain blissfully unaware of payload, 
> how, pray tell, do you deal with the challenge of popular and repetitive 
> content?

Neutral interdomain single-source multicast? I realize there is some 
technical and financial engineering required to make it work.

Or maybe we can convince content owners to use HTTP with proper cache 
headers instead of uncacheable protocols like RTMP or undocumented 
proprietary P2P protocols (e.g. Octoshape).

Barry Gold wrote:

> If it is illegal (or even immoral or
> fattening) for ISPs to look any further than the IP header when handling
> packets, then a lot of strategies like caching cannot be implemented.

Voluntary caching doesn't require DPI. I don't like DPI-based 
"transparent" caching because it often breaks things; in fact, I doubt 
that it can ever be made completely reliable given all the weird 
protocol tunneling out there.

Here are some examples of voluntary caching:
http://en.wikipedia.org/wiki/Web_Proxy_Autodiscovery_Protocol
http://bittorrent.org/beps/bep_0022.html

> I suppose, really, we should add a bit to the header saying whether
> intermediaries are allowed to look inside the packet.

I think that's called IPSec. :-) If you tried to seriously propose an 
"end-to-end bit" in the IETF, they'd just tell you that *every* packet 
should be treated that way and thus no bit is required.

Wes Felter - wesley@felter.org