[REBOL] no networking in experimental BeOS REBOL? Re:(3)
From: holger:rebol at: 15-Aug-2000 16:51
On Tue, Aug 15, 2000 at 05:14:21PM -0400, [ssayer--acuson--com] wrote:
> Similarly, I also have been unable to do any networking with the
> 0.10.24.4.2 (Linux x86 glibc 6) and the 0.10.25.3.1 (Win32) REBOL/View
> products myself on a couple of different machines (both with proxy & no
> proxy configs) which run network code with the release versions just fine
> (with the exception of the new tcp interface that I can't get to send or
> receive data connecting even to a telnet port 23 in the manner suggested
> by the Net docs). I haven't yet looked at it hard enough to tell if it's
> successfully opening sockets or not (kinda assumed it was known to be
> broken) but I will now.
There are no new known problems (known to us, anyway) in the networking
code in any of the experimental versions, with the exception of BeOS.
The problem with the BeOS version is caused by a bug in BeOS: the apparently
working non-blocking connect() function in BeOS actually has race conditions
that are not reproducable on our machines, but happen for some users. We will
switch back to normal non-blocking connect()s in the next experimental build
(with the unpleasant side effect that you cannot interrupt a connection in
progress with the Escape key -- same behavior as REBOL 2.2 everywhere and
REBOL 2.3 BeOS).
If you find any other reproducable problems please send reports to helpdesk,
and if possible provide examples we can run on our machines.
Regarding connections to telnet port 23: this is probably a bad example,
because telnet (and rlogin) are exceptions in that they make use of
TCP urgent mode
, so a fully compliant telnet client cannot be written
in most high-level languages. It requires getting "down and dirty" into
the depths of the TCP C socket API... Try connecting to any other server
(smtp, nntp, http or anything else).
--
Holger Kruse
[holger--rebol--com]
