> So after seeing this showmanship(kinda like a kid
> and a fast car) I shouted out that I can't get a 200 MB
> file from box A to box B ten feet away in less than
> 5 minutes. Thats with 100MB ethernet.
>
> Anyways, after making myself sound like a complete idiot,
> (and thankfully Ben diverted attention away from me)
no, you didn't sound like an idiot. :) sorry I broke off the conversation
with you in the middle of trying to solve your problem... I really did mean
to get back to you, but wanted to listen to the point that Ben was making at
the moment (since he'd made some progress with building the kernel).
> So why did they implement NFS using UDP? This just seems
> goofy to me. Theres an option to run NFS on TCP,
> but it's not the default.
NFS was originally designed to run over slow, possibly
high-latency, unreliable links (i.e. the Internet). As such, it may have
seemed more sensible to to transmission control at the application level,
rather than the transport level. (the application knows what it's doing, and
what situation it's in, better than the generalized transport layer does).
> I'll try out this netcat thing soon, as it sounds like
> a hip tool.
it's really simple, and really versatile. the man page pretty much explains
it all. ('man nc').
there are other, related projects like cryptcat (an encrypted netcat); and
one called firehose, which lets you trunk multiple connections together
easily. (so you can aggregate a GigE link, a 100MBps link, and a PPP link
together; at least in theory).
Carl Soderstrom.
--
Systems Administrator
Real-Time Enterprises
www.real-time.com