jas@proteon.com (John A. Shriver) (02/12/91)
In SunOS 4.1.1, udp_cksum is still off. The only progress towards any
sort of intelligence on Sun's part is that they finally *document* it,
and even cover how to set it by editing one C file when building a
kernel. And it finally does affect NFS! (It never did in SunOS 3.X!)
We've recently had serious NFS file corruption problems because of the
absolute impossibility of enabling NFS/UDP checksums on SunOS 3.5.2,
which we still use on a key development machine. (It was due to a bug
in a development router corrupting packets.) At least when I upgrade
it (someday) to SunOS 4.1.1, I'll get checksums.
Presumably, Sun chooses to ignore RFC 1122 (Requirements for Internet
Hosts -- Communication Layers). It is dated October 1989, so I
seriously doubt that SunOS 4.1.1 had a feature freeze that long ago.
(However, we have heard that the path from the TCP/IP group at Sun to
SunOS is very long, and fraught with delays, so it may well be that
the fix has been in the pipeline for a year. If that is the case,
kudos to the TCP/IP group, and would the SunOS group get off their
tails!) It would be nice to hope that SVR4 will conform to the host
requirements RFC's...
To re-quote the section of UDP checksums:
4.1.3.4 UDP Checksums
A host MUST implement the facility to generate and validate
UDP checksums. An application MAY optionally be able to
control whether a UDP checksum will be generated, but it
MUST default to checksumming on.
If a UDP datagram is received with a checksum that is non-
zero and invalid, UDP MUST silently discard the datagram.
An application MAY optionally be able to control whether UDP
datagrams without checksums should be discarded or passed to
the application.
I have no idea if Sun would reply to a bid that required RFC 1122
conformance and claim conformance. They don't conform.