Last August, there was a "Boston Tea Party" in which a number of
institutions connected to a Croquet party using a special version of
Jasmine (the "prerelease" of Croquet). This experiment used a
version of the hole-punching techniques, as implemented by David
Reed. I think there were more than a dozen machines connected.
Since then, we have used this code on our own experimental versions.
(
http://www.wetmachine.com/itf/item/429).
We field tested this in a class in a lab on campus, in a project
involving administrators on campus, and at our homes.
It worked great in the lab, crossing several LANs here. But the
results outside the lab were spotty. Even though Ford's 'natcheck'
test worked, other factors got in the way. Personal firewalls on
machines that closed ports, certainly. Port-blocking at various
network boxes, maybe. We did not have the budget to investigate further.
As a result, we (at the University of Wisconsin) are concentrating on
more traditional access: a Croquet router instance on a known port on
a public fixed IP address provides routing for parties whose members
may themselves be behind firewalls, NAT, etc. This is working much
better. I still believe that there is tremendous value in having
users spontaneously create their own routers, any time, any place,
for use in their own ad-hoc worlds. But this is not at the top of
our immediately funded list.
Note: The Ford paper describes how the TCP wire protocol does not
prohibit the activities required for hole-punching, but that the
system API's available to applications do not facilitate them. David
had to write extensions to the Squeak VM in order to allow this, and
these extensions did not make it into the current SDK.
Howard Stearns
Croquet Lead Developer,
http://opencroquet.orgDoIT Academic Technology,
http://www.wisc.edu/academictechUniversity of Wisconsin-Madison, 1301 University Avenue, Madison, WI
53715
+1-608-262-3724
On Apr 24, 2006, at 2:43 PM, Doug Jones wrote:
> It seems likely that people will continue to ask about connecting
> Croquet clients across the Internet. (I myself have several
> friends who are intrigued by Croquet, and that is always the first
> question they ask. They all have broadband at home, and without
> exception they all have broadband routers, and therefore are behind
> NAT.)
>
> A relevant paper was presented last year at USENIX. It was
> formerly available only to USENIX members, but now anybody can see it:
>
>
http://www.usenix.org/publications/library/proceedings/usenix05/
> tech/general/ford.html
>
>
>
>
Locked
