On Thu, Mar 13, 2003 at 06:31:24PM +0000, Timm Murray wrote: > > On Wed, Mar 12, 2003 at 06:32:04PM -0800, Ian Clarke wrote: > > > > Doing this ourselves is a waste of time. We'd be better off with a > > > > second-generation IP protocol like SCTP or RUDP. The former is a TCP > > > > replacement, the latter is a reliable UDP protocol layered over standard > > > > UDP. > > > > > > But how well does it support efficient fallback to TCP in the event of > > > UDP failure? How well can it disguise the fact that we are a freenet > > > node? > > > > Well, what are the cases where a UDP transmission would fail, but a TCP > > connection would succeed? Its not a firewall, as we'd already be on a > > port that a firewall would probably block. If a hole is opened for it, > > then one can just as easily open the UDP port. SCTP is a different case > > altogether, as its not built ontop of UDP anyway. > > > > How well can we disguise the fact that we're a freenet node over UDP? > > I think what Ian was getting at was that if Freenet uses SCTP/RUDP, and few other > applications use those protocols, then Freenet traffic will stick out to anyone with > a sniffer. UDP is fairly common (especially with the popularity of > multiplayer games), so it's not a big deal. > > Of course, this won't be a problem if/when SCTP/RUDP becomes popular.
RUDP *is* UDP; its a protocol for using UDP reliably. Even so, using
UDP and our own scheme makes you just as easily detectable because of
the way we'll be using it (Thats a freenet node because of what it did
when packets were not acknowledged, and how hard it kept trying, eg).
Scott
pgp00000.pgp
Description: PGP signature
