The present page I'm worrying about is
http://eddie.mit.edu/cevat/report/models.html
which is the "long-range models" section. I differ strongly with one central
premise to this page: that at any one time there is a single model whereby
networked computing (or other technologies) are deployed, and a new model
necessarily displaces the old. Instead, I think that new models permit change
in old models that was heretofore impossible, and that we must attend to
parallel changes in different computing models rather than oversimplify the
situation at the expense of educational value.
The paragraph that sets me off is the fifth, which begins "Recently the MIT
network has been extended..." and goes on to say this:
"The students will own the computers, and MIT
will supply the network and the necessary infrastructure,
including mail servers, print servers, Web servers, permanent
data storage, and some personal computers on
campus for occasional use by students during the day."
Now as you know I think it's important and inevitable that MIT move toward
expecting (if not requiring) all undergraduates to own personal computers of
some reasonable level (although this is as controversial among
academic-computing staff in IS as it is among faculty and students), and that
once this happens a great deal of what now goes on in public clusters
(especially the activities I generally group under "Commons" --
communications, information retrieval, and so forth) should move to those
personal computers from MIT-owned networked workstations. But the draft report
implies that this will free MIT from the need to provide public workstations,
and I disagree strongly with this.
The current demand from faculty for new facilities involves workstations two
notches above those students typically buy: students buy $2k personal
computers, the typical public-cluster machine is a $5k workstation, and the
machine in greatest demand from faculty for new initiatives is a $10k
graphical workstation capable of doing modest visualization stuff. (There's
demand for higher levels too, but I'm ignoring that for the moment).
There's an enormous Institute educational investment in applications on the
$5k machines, which will only gradually (and never completely) move to more
heterogeneous student-owned machines. We can do some disinvestment in $5k
machines as student-owned machines begin carrying Commons load, but in my view
that disinvestment will be more than offset by new investment in more capable
machines. That is, I don't think that the *number* of public workstation seats
is going to diminish dramatically at MIT as personal networked computers
proliferate; rather, I think their distribution -- as to both capability and
deployment -- is going to change. If the number of seats does decline, then so
will the educational effectiveness of MIT academic computing.
The capabilities of the $2k, $5k, and $10k machines will increase over time,
of course (and the specific prices, but not their relationship, may too). But
I believe that history teaches us today's basic trend -- $2k machines take on
some work previously done by $10k machines, some public $2k and $5k machines
remain necessary for on-campus use and platform sufficiency for specific
applications, and savings on $2k and $5k machines go to $10k machines -- will
remain the case for a long time.
Even granting this point, which some committee members may be unwilling to do,
there's the question Hal raised about whether $10k machines (what I was then
calling "specialized" machines, by which I meant with *additional* special
capabilities rather than narrow ones) should be deployed centrally or
departmentally. In my view there are a few departments with the infrastructure
and critical mass of educational use sufficient to justify departmental
deployment (the 6.001 lab is a good example, but there are very few 6.001's at
MIT), and departmental deployment may make sense in these departments, but for
the most part I think the current central model will continue to serve MIT
best.
I don't think the Committee will be able to resolve these points (either the
single path versus the distributional shift, or the central versus
departmental deployments), but I certainly think we should describe the
disagreements and present the various arguments.
More thoughts as I delve deeper. I'm especially concerned that the
oversimplification I see in this section not be reflected in the corresponding
recommendations -- that is, if we recommend that MIT require computer
ownership by students, we have to be careful not to assume that the costs of
this can be recovered from the public environment without educational harm.
gj
e40-359a/MIT/Cambridge MA 02139
voice: (617) 253-3712
fax: (617) 258-8736
url: http://web.mit.edu/gjackson/www/home.html
key: pgp@pgp.mit.edu