[1554] in Release_7.7_team
notes, 12/1
daemon@ATHENA.MIT.EDU (Mike Barker)
Sat Dec 5 19:43:34 1998
Date: Sat, 05 Dec 1998 19:32:02 -0500
To: release-team@MIT.EDU
From: Mike Barker <mbarker@MIT.EDU>
Attending: Jonathon Weiss, Miki Lusztig, Robert Basch, Abby Fox, Mike
Barker, Bill Cattey, Dan Winship, and a masked stranger
1. The "evil SGI attacks and kills AFS servers" problem. It is believed
that
this occurs with Netscape running and expired tokens. There is an Elmer
change which would cause tokens to be deleted one minute before they would
expire. Although we do not known for certain if this will stop the evil SGI
from attacking AFS servers, it seems to be worth trying, at least to see if
it changes the problem.
We're seeing the problem infrequently (once maybe twice a month) but
depending on which server it brings down, it may be very visible.
Given the pressure to use systems at the end of term, we would not expect
that a system would be left alone to do this.
Our suggestion: hold off until after finals.
2. As an alternative, the release team recommends that we desupport the Web.
3. OS patch release. The idea is to bring the operating systems closer to
current patch levels sometime in January.
For the Solaris systems, Miki thinks this would be a good idea. She says it
will take her about two days, and Greg would then finish. The release team
asks Greg to coordinate with Miki and make this happen.
For the SGI, this is not such a good idea. They do not have recommended or
security patch lists, relying largely on "if you have this problem, install
these patches" and large interdependent patch sets. Bob will look into
whether there are patches that we would like to have.
One upgrade we may do is for the new AFS client, which has the correction
for executable corruption.
4. The problem in 4-035 has not been resolved. Experimentation has shown
that Bob's office has one drop that appears to have network loss, while
another drop appears to work perfectly. This may make analysis of the
problem somewhat easier, although the biosym loading problem does not seem
to occur.
5. We are still explaining what we are trying to do with the SGI inst
program to SGI.
6. The problem with network booting requiring hexadecimal entry on Suns
seems to have re-surfaced on the enterprise 450 and 250 systems. Bill is
working with Sun to make sure the fix is applied, and to help them learn how
to apply patches across a source tree so that one fix takes care of all
occurrences.