24C3: Current events in Tor development

[coderman]

On Jan 9, 2008 5:14 AM, Roger Dingledine <arma at mit.edu> wrote:
> ...
> My talk at 24C3 in Berlin is now up on the web in various formats.

hi Roger,

this looks like an interesting talk; i wish i could have seen it in
person.  some comments...


> "Theory: Tor is slow because a handful of people are running file­sharing apps on it"

is there any research / observations to support this claim?  i'm
curious.  (i think Tor will always be "slow" compared to direct
traffic.  this is a feature, not a bug, and exists independent of any
particular transport protocols...)


> "Vidalia should learn how to talk UPNP to routers"

i wish i could add a prerequisite before this item: "Implement
datagram based transport" :)  commodity routers even now have problems
with NAPT table entries for busy Tor / other network apps.  a UDP
transport would open up effective new NAT busting abilities and bypass
some of the windows / weak router NAT table conn tracking issues that
make scaling the existing TCP implementation difficult.  i know the
DTLS proposal hasn't seen any attention lately.  is this potentially
going to resume sometime, or is an entirely new proposal needed?  (the
UDP section mentions more research / hacking, maybe even both? :)


best regards,