[ On Thu, March  2, 1995 at 21:25:03 (-0800), Ted Lemon wrote: ]
> Subject: Re: rdist vs. sup 
>
> ...that you see the occasional corrupt packet slip through the header
> checksum.   Nonetheless, since a lot of us NetBSD hackers are in fact
> on the other end of slip or ppp connections, a 32-bit CRC in sup would
> be a big win.

That's still fixing the wrong thing.  If TCP/IP transports *guarantee* a
clean transmission circuit, then they should provide it.

If you're getting error rates like that, then you've *probably* got
hardware problems -- at least that's the level I'd start working on it
at.  Once you can prove that your error correcting modems are indeed
correcting errors, *and* you've got the same clean data going in and out
of your system, then you'll probably have solved the problem.

If you're running over a non-error correcting channel, or you've got
such a terrible async subsystem in your hardware that you're going to be
dropping packets unless you run at 2400bps, then you should find out why
TCP isn't correcting these errors.

It kind'a bugs me a great deal that the networking code in NetBSD *can*
allow errors to creep into a TCP circuit!

-- 
						Greg A. Woods

+1 416 443-1734			VE3TCP		robohack!woods
Planix, Inc. <woods@planix.com>; UniForum Canada <woods@uniforum.ca>