[6184] in Release_7.7_team
On the subject of de-supporting Athena 10 sooner rather than later.
daemon@ATHENA.MIT.EDU (William Cattey)
Thu Jan 29 09:18:09 2009
Mime-Version: 1.0 (Apple Message framework v753.1)
Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed
Message-Id: <11AFAAEE-172D-48ED-8026-1A752659EBDE@mit.edu>
Cc: release-team@mit.edu
Content-Transfer-Encoding: 7bit
From: William Cattey <wdc@MIT.EDU>
Date: Thu, 29 Jan 2009 09:17:24 -0500
To: "gettes@mit.edu Gettes" <gettes@mit.edu>
X-Spam-Flag: NO
X-Spam-Score: 0.00
Mike,
You asked for some thoughts on de-supporting Athena 9 sooner rather
than later.
I brought up the question at Release Team, and we spent some time
discussing it. As a consequence, we've adjusted our "Should we fix
it or leave it broken?" filters to have a stronger preference to
leave things broken.
As I've continued to meditate on the relationship between Athena 9
and Athena 10, I've realized that I and the other Release Team
members have been transitioning to trusting that Athena 10 really
will be a product, and really will be the successor to Athena 9. The
clearest example of this is in our approach to Firefox 3. We'd
assumed that integrating Firefox 3 into Athena 9 was just, "the thing
to do." It was becoming universally available on Windows, Mac, non-
Athena Red Hat, and so Athena 9 needed to keep up. But we didn't
really discuss whether it was better to leave that new feature as
incentive to migrate to Athena 10. I think we may have unconsciously
considered "transition to a different browser," as an obstacle to
adopting Athena 10, and got into the habit of making the transition
to Athena 10 from Athena 9 as invisible as possible.
Right now our target date for production quality Athena 10 is July 1,
2009. We got to that date by our traditional practice, "No major
changes to Athena during the school year." But perhaps we can be
more aggressive. We got to that practice because we'd taken early
criticism for breaking things in ways that prevented vocal faculty
members from teaching class with Athena. Maybe these days we can
target larger chunks of the Athena environment to transition to
Athena 10 during the school year, as long as we publicize where
people can go if they need the old system.
As I said at the DSPS meeting, Athena 10 is not ready for production
use. In fact, we're late to go live with the Early Cluster. Every
day Jonathan Reed ask each other, "When can we go live with Athena 10
Early Cluster?"
To go live in a Production Ready Athena 10 we need to resolve a bunch
of issues that we are tracking in Jira. We're at the, "I thought we
had it all sorted out! OOPS!" phase of early testing. I and the
other members of Athena Release team review the issues asking, which
issues are big enough that they'll discourage trust in Athena 10,
which issues can be put off, and which issues can be shed. To be
honest, we're all out of practice going live with something this
different from what we've done before, so we're being a bit cautious,
and our estimates on, "When will you be done?" are off. But we're
getting better as we gain experience.
Two important things need to be addressed for production quality go-
live, that we have been putting off because the work needs to be done
outside ISDA:
1. Documentation
2. Dialup/Timesharing Athena.
Documentation:
Lee Ridgway and Heather Anne Harrison know they're the ones to work
on, "A fairly large project going through Athena documentation,
throwing away the obsolete stuff, and updating the important stuff to
align with Athena 10." Today their priorities center not around
Athena but around the new Drupal-based IS&T Web Site. With a target
Production Ready go-live date of July 1, we all felt comfortable with
where Athena 10 sat in their priorities. Additionally, Jonathan Reed
has taken the initiative to get some of the critical documentation
ready in time for Early Cluster Release.
I will have a chat with Lee and Heather Anne to assess the impact of
trying to get the critical Athena 10 documentation ready sooner than
July 1. Heather Anne and I have found ways to leverage student and
volunteer labor before, so I am optimistic we can do better than July
1, but we need to think together and do some new planning to estimate
how much better.
Dialup/Timesharing Athena
I have allowed my understanding of the scope and value of the Athena
Dialup/Timesharing service to fall badly out of date. In some very
preliminary conversations with Jonathan Reed, I've learned that it is
regarded as an important service. The value, simply stated is, for
users who don't have their own private Athena system, they can sign
onto dialup.athena.mit.edu and conveniently get software and services
available in Athena clusters. More importantly, MIT faculty, staff,
and students who are off campus make use of this service.
Athena Dialup is on Solaris systems that have had additional security
patches made.
Preliminary discussions with (I forget who specifically I spoke to)
Athena Server Operations gave the answer that they were all flat out
in the migration to the new VM-CoLo based service model, and REALLY
didn't want to think about re-creating the Athena Dialup service on
Linux for a while.
The Athena Dialup service is a driver for keeping Solaris Security
patches up to date and keeping Solaris Athena desktop applications up
to date. If there is significant delay in deploying an Athena 10-
based Dialup service, we need to choose among:
1. Continuing to do work to integrate Solaris updates. (Athena
Dialup uses *Athena* Solaris as its basis.)
2. Continuing to do work creating Athena 9 updates to support Athena
Dialup
3. Offering a moribund Athena 9 dialup service for a period of time.
4. Shutting down Athena Dialup for a period of time. (I can already
hear Jonathan Reed and Oliver Thomas emphatically saying this is an
unacceptable choice.)
5. Expediting creation of an Athena 10 Dialup service. SIPB has a
prototype, they call it Linerva.
I should also check in with Jonathon Weiss to learn if there are
other services run by IS&T that require Athena Solaris that might
drive requests to do Solaris or Athena 9 work.
I think Athena Release Team is in the process of shifting mindset to,
"Leave it old or broken in Athena 9 as incentive to go to Athena
10." We did take perhaps a bit too much new and useful stuff into
Athena 9 over the past year. Sorry. We're changing.
I'm CCing Release Team to let them know what I've said. Release team
members are welcome to correct me where I've gotten it wrong or to
add in additional missing aspects of what we would have to do to
IMPLEMENT, "Cut over to Athena 10 sooner rather than later."
-Bill
