Meeting Summary
===============

The DCE Task Force Working Group Chairs met last Thursday, August 25,
from 2:00 - 3:30 in the Ben Franklin Room of Houston Hall.

We reviewed and discussed what we learned from Ted Hanss' visit, as
well as Jonathan Kline's presentation on IBM's DCE plans. (Details
below.)

I reported briefly on my week in Transarc's courses on DCE Secure Core
and DFS System Administration. We agreed that I should relate the
lessons I learned in a weekly, one hour course starting right after
the semester starts and continuing until we are finished. Unless
someone raises compelling objections, the class will be held on
Wednesdays from 12-1 or 1:15.

We also discussed how to proceed from here, agreeing specifically to
make more formal use of the Information Technology Architecture
methodology now being used by the Network Architecture Task Force.
This consists of identifying four major component inputs into the
formulation of a new architecture:

1. Principles
2. Business Needs
3. Current Architecture
4. Technology Trends

Ted Hanss started the fourth endeavor, and Jonathan Kline from IBM
continued it. The working groups are eager for more vendor
presentations (like the NATF ones) from folks from OSF, HP, DEC, Sun,
Apple, Novell, MicroSoft, Transarc, and Oracle.  I was asked to
coordinate communications with these organizations so as to avoid
confusing them with a barrage of queries from working groups.

Findings from Ted Hanss' & Jonathan Kline's visits
==================================================

Among our findings are the following (some of which were posted to
upenn.net.dce by working group chairs):

1. This is a big effort. Michigan has been working on their DCE for
years, with a large staff, external funding, and a tradition of
home-grown, centralized computing, and they are only now really
rolling out the DCE to the campus at large.

2. The big advantage to end users will be increasingly ubiquitous
access to a consistent computing environment, more or less as
expected.

3. One person wondered if it wasn't too soon to pick specific
technologies.

4. It will take a long time to set things up, so the fact that many
products may not yet be ready for prime time is much less of a
problem.

5. Even the foundations, e.g., authentication services, are in
transition, and will be for many years.

6. Cooperating with vendors will be necessary to help direct product
development in ways that will meet our needs.

7. One person reported that he had been surprises by Ted's description
of Unix facilities that were less mature than expected, while PC and
Mac integration was more developed that expected. Granted, the Unix
stuff is much further along than the PC and Mac stuff (which is
typically gatewayed to hell and back), but not as much as expected.

8. Reports from the Share conference, discussions with Transarc
employees, and IBM's recent purchase of Transarc, all combined to
impress upon the group that IBM was serious about DCE and DFS.

9. The question was asked as to how many cells the University needs,
for which there is no technology independent answer. Using AFS would
require that we deploy cells for nearly every school, many
departments, and even centers, obviating the advantages of
single-signon authentication for anyone straddling a departmental
boundary. Under DCE, DFS allows distributed control of file systems,
which would greatly reduce the need for cells. Under DCE 1.2, reported
due to release from OSF next fall or winter, there is supposed to be
support for "hierarchical cells". Although none of us has seen any
details of what they are, Ted more or less said they would allow us to
match one hierarchical cell, administered much as the rest of Penn.

10. One person thought we needed to implement _both_ AFS and DCE/DFS
cells to gain the advantages on AFS' maturity, while anticipating the
dominance on DCE and DFS. The DCE Secure Core has the potential for
use by Project Cornerstone's telnet and Oracle client/server
applications. Oracle is reportedly in excellent position to deliver
DCE security-enabled applications. We suspect however that there is
nothing in our contract with them addressing such needs.

It was noted that Ted Hanss recommended going to DFS by way of AFS if
we were starting today, and suggested that a year from now we could
either do that, or go directly to DFS, and that in two years he would
probably only recommend DFS.  Jonathan Kline on the otherhand
recommended going straight to DFS.

The suggestion that we should do both DCE/DFS and AFS was discussed at
length, and the person responsible (me) eventually escaped with his
life :-), as the group concluded that we should focus on DCE and DFS.
Among the telling points were that no one thought that joining the
existing interorganizational AFS tree, and other advantages resulting
from AFS' greater maturity, were worth the expected difficulties
converting from NFS and then again to DFS. This was especially true in
that AFS requires a wholly centralized approach to cell (aka realm)
and file server management that is untenable at every level of the
University.

Next Meeting (**** This week!****)
==================================

Our next meeting will be this week, on Thursday, September 1, from
2:00-3:30 in the Gates Room in Van Pelt Library.

The meetings after that will be September 22 and October 6.

(Although we are trying to meet on the first and third Thursdays, I
have not yet found a new home on the fourth Thursdays for the
NW-Admins Team, which presently meets on the thirds. Hopefully I will
resolve this during September to take effect in October. Sorry for the
inconvenience.)

Please feel free to send corrections and ask questions.

-Chris