| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
Date: Wed, 18 Dec 1996 03:11:45 -0500 (EST) To: Mike Whitson <mwhitson@MIT.EDU> Cc: Bill Cattey <wdc@MIT.EDU>, Mike Barker <mbarker@MIT.EDU>, hotline@MIT.EDU In-Reply-To: "[36095] in Hotline Meeting" From: John Hawkinson <jhawk@MIT.EDU> (belated, sorry) > The cost of living with the broken proms for now is minimal. Mark > Virtue and I have already created a workaround for the new bootp > server which involves turning off MTU discovery in its IP driver. It > caused us about 6 hours of headaches trying to find the problem, but > that's done with. The real concern is that it causes an obsure > dependency in our SGI boot service on an obsure Solaris IP driver > configuration option. When we next upgrade the boot server to a > different platform, who knows if this will continue to be a problem? In case you care (probably you do not), could could deal with this problem by consing up the packet in question using Solaris' interface for sourcing layer-two frames (DLPI). This would be about a day's work and may not be worth doing, as not very many other operating systems support DLPI as a mechanism for sourcing raw frames. Nevertheless, once you had written such software using the DLPI interface it would pretty straightforward to port it to any other interface that supports this functionality (eg: NIT, bpf, or SGI's snoop interface). It would be reasonable to expect any other modern operating system you ran the daemon on to source UDP packets with the DF bit set. --jhawk
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |