R. Clayton (rclayton@clayton.cs.monmouth.edu)
(no date)
The answer to the pending question of using connect in UDP servers seems to be
in the Snader book (Pages 131 and 132).
I was under the impression that the open question about connect was whether
or not a connected udp socket could be reconnected to some other address. In
particular, is the following legal under the Sockets API:
s = socket(udp);
bind(s, ...);
connect(s, ipa1, pno1);
a bunch of io over s;
connect(s, ipa2, pno2);
a bunch more io over s;
This server is designed to be launched by inetd. That brings up another topic
"the architecture of inetd launched servers", which seem to be a little
different than the ones we discussed in the class until now.
Yes; inetd servers have a different structure than do stand-alone servers.
Essentially, inetd servers deal with standard io (std-in, std-out, and
std-err) instead of sockets. We're scheduled to take a look at them on 27-29
March.
Can we develop a toolkit based on the Snader book source code after we
understand it? OR Do we need to develop the code on our own?
I have enough faith in the unstraightforwardness of sockets programming that
I consider these two options to be the same. Even if you copy the code right
out of the book, you will encounter enough problems that you will be forced
to go back and understand the code in order to fix it. This is particularly
true of you're going to try cross-platform development, and especially true
of the platforms are windows and unix.
I would rather encourage everyone to develop their own toolkit, however they
did it. Prospective (or actual) project partners could then get together to
review each others' work.
This archive was generated by hypermail 2.0b3 on Sun May 06 2001 - 20:30:05 EDT