[1108] in SIPB_Linux_Development
SIPB-Athena versioning (was Re: i386_linux1 locker)
daemon@ATHENA.MIT.EDU (Greg Hudson)
Sat Oct 21 14:00:04 1995
To: Erik Nygren <nygren@MIT.EDU>
Cc: Greg Hudson <ghudson@MIT.EDU>, linux-help@MIT.EDU, linux-dev@MIT.EDU
Cc: netbsd-dev@MIT.EDU
In-Reply-To: Your message of "Sat, 21 Oct 1995 13:36:54 EDT."
<199510211737.NAA20572@foundation.mit.edu>
Date: Sat, 21 Oct 1995 13:59:23 EDT
From: Greg Hudson <ghudson@MIT.EDU>
I've added netbsd-dev to the cc line and dropped davidz.
> Is anyone relying on that directory at this time? I might want to
> do some stuff to clean it up but I don't want to hose anyone who's
> using it.
I don't have any reason to believe anyone is relying on it; if they
are, they're probably losing pretty badly. (They're certainly going
to lose badly now, since I've just moved everything.)
Before anyone mucks too much with system/i386_linux1, I'd like to
correct an error that was made with /afs/sipb/system/i386_nbsd1;
specifically, the lack of a versioning structure underneath it. Under
/afs/athena/system/*, you have srvd.* directories and a link srvd to
the most current directory.
In order to have a similar structure, we need a version numbering
scheme for SIPB releases. After some reflection, I think we want
version numbers of the form:
srvd.77.0A Athena 77-based, testing prior to version 1
srvd.77.1 Athena 77-based, version 1
srvd.80.0A Athena 80-based, testing prior to version 1
srvd.80.1 Athena 80-based, version 1
srvd.80.1A Athena 80-based, testing prior to version 2
srvd.80.2 Athena 80-based, version 2
Through some small stretch of the imagination, I'm calling our current
source tree "Athena 77-based". (In six months to a year, after I
integrate most of our changes back into the Athena source tree, I
would like to start moving towards something more like a "SIPB version
of the Athena source tree", so the bases will become a little more
clear.) srvd should normally point to the last non-testing version,
although when we only have a testing version available (e.g. for Linux
right now), it will point to that testing version.
We actually have some disk space now so that we can have multiple
versions of parts of the system packs (/usr/andrew and /usr/X11R6 can
probably be the same across most versions), but we don't have enough
disk space to leave releases around forever. As such, we should
probably develop a system for keeping track of whether releases are
still being used. Something along the lines of: at system boot or
reactivate time, if /srvd is mounted remotely, send a UDP packet to
some SIPB service (which will not remember the from address)
containing /srvd/.rvdinfo. I don't see this being a privacy
violation, since such machines are already sending lots of UDP packets
to SIPB AFS servers in the course of retrieving files.
I've moved /afs/sipb/system/i386_linux1/usr to
/afs/sipb/system/i386_linux1/srvd.77.0A and made an srvd symlink to
srvd.77.0A.
For NetBSD, I will start imposing this structure after the NetBSD 1.1
release while I'm rebuilding our /usr/athena tree from scratch. I'll
leave behind the old usr/athena, leave compatibility symlinks for
usr/{andrew,X11R6} and emul/linux, and probably strip out the rest for
cleanliness.
If anyone has differences of opinions with regards to what I've said,
please say so.
