[35] in Info-AFS_Redistribution
Re: More on AFS over slip - getting closer
daemon@ATHENA.MIT.EDU (daemon@ATHENA.MIT.EDU)
Wed Dec 26 17:13:58 1990
Date: Wed, 26 Dec 1990 16:36:10 -0500 (EST)
From: Craig_Everhart@transarc.com
To: Info-AFS@transarc.com
In-Reply-To: <4bS_r2M91E5zEwH1tu@rchland.ibm.com>
I'm not sure what commentary Bob O. is looking for. I, for one, would
be happy to try to incorporate Rx fixes into the current product.
I've seen the rx.c stuff deciding on a packet size, and considering
whether a machine is ``local'' or not. I'm not sure what ``local''
predicate would be good to check.
There's apparently a new socket ioctl that allows a program to read the
interface's MTU value--the MTU in use on the connected medium. Using
that would probably be better yet, but I haven't been able to get the
ioctl to work on any kernel I've tried around here. Then again, inside
the kernel one could presumably just read the appropriate word from the
kernel data structure. If it's not difficult, it could be great.
A somewhat bigger black hole would be implementing the Path MTU
Discovery protocol.
A yet larger black hole would be incorporating slow-start kinds of
protocols into Rx (if in fact doing so would be a good idea).
Does that answer your questions?
Craig
