[5686] in Release_7.7_team
The future of Athena
daemon@ATHENA.MIT.EDU (Greg Hudson)
Thu Jan 11 18:29:06 2007
Date: Thu, 11 Jan 2007 18:29:03 -0500
Message-Id: <200701112329.l0BNT3vb023969@equal-rites.mit.edu>
From: Greg Hudson <ghudson@MIT.EDU>
To: release-team@MIT.EDU
X-Spam-Flag: NO
X-Spam-Score: 0.00
Disclaimer: These opinions are my own. They are not necessarily the
opinions of IS&T's leadership. This missive hasn't been reviewed in
advance by anyone, and I'm going out on a limb by sending it to a
mailing list with a public archive. Please don't take anything in
here as official statements or commitments by MIT or IS&T.
Also, it's kind of long.
1. Where we are
When Athena was first introduced, computers were expensive and most
students didn't have them. Athena provided the only universally
available computing environment. That stopped being the case maybe
ten years ago. Athena continued to provide value as a primary
computing environment for some years after that because it had a
superior story in the areas of security and software deployment, but
that stopped being the case several years ago as well.
Athena today is still a widely-used resource, but very few people call
it home. About of half the campus population uses it during a given
month, but the median number of monthly logins hovers around nine, or
once every three days. People use it to retrieve or publish shared
information in AFS, to run specific pieces of software, or to access
MIT infrastructure from a convenient location away from home without
carrying around a laptop.
So Athena rests in the awkward position of many legacy services: too
valuble to easily kill, but not exciting enough to invest in.
2. Where we're going
For the past year or so we've been focusing on how to keep Athena
going with a minimum of effort. Current plans to this end include:
* Reduce the platform count from two to one, by phasing out Suns.
Athena 9.4 will likely be the last release for Solaris on the
desktop, and may be the last release in general depending on the
needs of Server Operations.
* Perform full releases only when something gives us a kick in the
pants to do so, instead of doing them proactively each year. A
year ago I thought we would probably get a kick in the pants for
summer 2007, but right now it appears we can wait until at least
summer 2008.
* Select operating system bases with long term support commitments
from their maintainers. RHEL 4 is an example of such; SLED 10 and
Ubuntu 6.06 are other examples. Ubuntu 6.10, Fedora Core
releases, and OpenSUSE releases are counterexamples.
* The next time we do a full release, take some effort to stop
building 90% of the source tree and use native Linux distribution
components instead. We've done some prototyping work to this end,
though I've suspended the completion of that work until closer to
the date of an expected full release.
* Possibly outsource maintenance between full releases to
contractors. I don't personally relish doing the work required to
make that possible, but it may happen anyway.
3. What's cool about Athena
Why is Athena so widely used when almost everyone has a desktop
environment they like better? I don't really know for sure, but I do
know what I think is cool about Athena:
* It's really easy to store and share files on Athena. The files
you store are backed up for you and everyone else on Athena can
see them if you want.
* It's very easy to access MIT infrastructure from Athena. Services
like email and instant messaging are wired right into the default
environment, and the default web browser is preloaded with an MIT
certificate.
* It's customizable. If you don't like the way it looks or works by
default, you can make it work a different way.
* It's user-extensible. Anyone can install software on Athena, and
with a little effort can even give it a semi-official-sounding
place to live.
* It's expansive. It serves such a variety of functions that it
never feels fully explored.
* It provides a rich scripting environment which lets power users
piece together components into their own mini-applications.
* It's location-independent. The particular machine you're using is
unimportant most of the time, and you get (roughly) the same
experience no matter which one you use.
* It's universal, sort of. Everyone you know at MIT can access it,
although many of them might not be excited about doing so.
A lot of these aspects are missing in commodity desktop environments,
but even combined, they aren't compelling enough to make people use
Athena over the desktop they're used to. It doesn't help that Athena
is intrinsically slow and based on unfamiliar, somewhat marginal
technology, but that's not the major issue.
4. Is there something equally cool that people actually want?
If people already have desktops, is there anything they want from a
central IT organization that can be as expansive and cool as Athena
is? I think there is, and it's web hosting. A computing presence and
resource that sticks around when you close your laptop, and can be
accessed (via a web browser) from random kiosk machines anywhere.
Done well, an IS&T web hosting service can have all of the cool
properties I listed above, and can interface with existing and new MIT
infrastructure easily. It can be user-friendly enough to appeal to a
wide audience (anyone who wants any kind of web presence such as a
wiki or blog) and flexible enough to appeal to power users who can
extend the environment like they have Athena. Linux is not a marginal
technology for web hosting, and a kernel-loaded filesystem module like
AFS won't be necessary to make it work.
People in SIPB realized this over two years ago, of course, and
started providing scripts.mit.edu. I think it's done an excellent job
within the constraints SIPB has, but that IS&T can do something much
more ambitious.
So far this is just an idea in my head which I've shared with a few
other people, not a plan or an organizational commitment. It's
possible that the idea will crash and burn, or that it will change
form or focus before it reaches a pilot stage. It also probably won't
reach its full potential without improvements to other areas of MIT
infrastructure, such as file sharing and integrated communications, so
it may wind up getting bundled together with something else.
This is not an idea for an Athena replacement--as noted above, Athena
is too valuble to easily kill, and this wouldn't really replace its
value as a resource. This is more of an idea for a spiritual
follow-on.
