[Previous] [Next] [Index]
[Thread]
Re: Hacking a Personal Computer via E-mail
> Well...if it weren't for little details like timeout values, I suppo
se....
>
> You do realize that you can't even run TCP over two satelite hops be
cause the latency is greater than the timeouts, don't you?
>
as said, i nerver did it, but i was told by private mail it's
been done.
Whether or not it's feasible depends on the TCP implementation. While
I've certainly seen PC stacks that give up after a few seconds, modern
UNIX-based TCPs can last a fair number of minutes. RFC 1122 requires
a timeout of at least 100 seconds before giving up. According to
TCP/IP Illustrated Vol II, 4.4bsd's TCP waits at least 500 seconds
before declaring a connection dead.
Btw -- I know someone who's implemented it. ``Lousy latencies but
a big MTU....''
--Steve Bellovin