[934] in NetBSD-Development
8/2/95 meeting notes
daemon@ATHENA.MIT.EDU (ghudson@MIT.EDU)
Wed Aug 2 19:54:37 1995
From: ghudson@MIT.EDU
Date: Wed, 2 Aug 1995 19:53:24 -0400
To: pc-dialup@MIT.EDU
Administrative note: netbsd-dev will be removed from this list and
replaced with a private discuss archive in the near future.
Here are the issues that came up at the meeting:
* Source and binary trees
It will take 300-500MB for a full set of source and binary
trees (not including build trees). Greg will set up the ops
source tree and help maintain it.
* Integration of NetBSD port into the Athena source tree
Greg is preparing a report on the NetBSD Athena port, and
may be doing rel-eng work for DCNS in the future. Craig is
very busy with 8.0 issues, so it will be at least three
months before we start integration.
Kerberos will probably not be integrated because we use CNS
Kerberos.
Most major changes are to the login code; Greg plans to have
the login utilities all use libAL (not much work remaining
at this point), which is a goal for the Athena port. xlogin
integration may be difficult because of the number of
changes on the Athena side.
* Access (.klogin) on the dialup
Anyone making positive contributions can be in the .klogin
file while this is a pilot project. When we announce this
as a service, we will have a testing machine for developers
to resolve problems.
* Hardware problems
- BIOS doesn't enable DMA bits, so we have a kludge to get
the ethernet card to work. Not an operational problem
right now.
- When video memory caching is enabled, the bus sometimes
hangs. Disabling video memory caching makes the problem
go away.
- The BIOS crashes (hangs) after starting to reboot. This
hasn't been fully investigated. This is going to be a
(minor) operational problem if it can't be worked around.
These problems make Zeos unimpressive (especially since they
shipped us a developmental BIOS), so we may want to consider
other vendors (e.g. Dell Dimensions) before we buy more
machines. Maybe we can get evaluation machines.
* Names for the dialup
Karl chose "ringworld" for the primary name; we'll let this
stand for now.
The macine will live in a netbsd.dialup pool during the
pilot project.
* Items to do before we into production
sendmail needs to send outgoing mail through mit.edu.
look at sendmail problem with users not in passwd file
(specifically, mail about vi recovery files).
snmpd support for load balancing - needs load, number of
users, and existence of /etc/nologin. CMU SNMPD has been
ported to NetBSD, or we can port the Ultrix version.
We'll probably use cheap video cards instead of serial
consoles, since serial consoles won't work if the CMOS
gets scrambled.
Kernel hacks for blocking X connections.
Make sure "cleanup" works properly when users are logged in.
* Security
No specific issues raised.
krb5 telnetd? Wait for things to stabilize; maybe we could
try it on pesto.
* motd should tell people service is experimental, and point
people at pc-dialup for problems.
* Mailing lists
Karl wants to use ringworld-engineers. Thump.
Charles suggests netbsd-cabal.
We decided to use pc-dialup with netbsd-dev taken off,
archived to a private discuss meeting.
* Location
Charles still needs physical access while kernel-hacking.
We'll move it elsewhere in the zone for now.
* Installation
Greg will work on an automatic installation procedure once
the source and binary trees are set up.
* Who can log in?
jweiss suggests restricting, but we decided that security is
probably at least as good under NetBSD as Ultrix, so we
should let everyone log in.
