Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!ux1.cso.uiuc.edu!ux1.cso.uiuc.edu!m.cs.uiuc.edu!p.cs.uiuc.edu!gillies
From: gillies@p.cs.uiuc.edu
Newsgroups: comp.edu
Subject: Re: CS Ph. D. program
Message-ID: <82400052@p.cs.uiuc.edu>
Date: 20 Oct 89 00:40:26 GMT
References: <160@seqp4.UUCP>
Lines: 40
Nf-ID: #R:seqp4.UUCP:160:p.cs.uiuc.edu:82400052:000:1872
Nf-From: p.cs.uiuc.edu!gillies    Oct 19 00:34:00 1989


> I am interested in a 'Systems' type program.  My main field of
> expertise is database systems.  I am also interested in distributed
> systems, parallel computing, and software engineering.  

"Systems" and "Software Engineering" are very interesting, but they're
tough areas to get a Ph.D. in.  If Illinois is a good barometer, do
not expect to make a large software project into your Ph.d.  The Ph.D.
is fundamentally different from the M.S. in this respect.  Also,
thesis in these areas takes THE LONGEST of nearly all the disciplines
of C.S. (except perhaps AI).

If you want to teach, it's hard to survive in the Systems area,
because during your first five years you have little manpower, you
probably need several dedicated machines and also need decent
low-level tools (both these facilities are hard to find at even the
hardware-rich schools).

There are a number of large, stagnant distributed systems projects
rotting at various schools.  Many of these systems are dormant, so
don't expect to contribute to these projects.  Make sure you
investigate the distributed systems projects carefully.

Perhaps you should exploit your interests in database and/or parallel
systems.  I know it's not on the Coast, but Illinois has several
parallel processing and supercomputing projects (the Cedar project,
the Tapestry hypercube OS project, and NCSA).

You can go to a school and become a coder on some high-profile
software system (X11, Ingres, 4.xBSD), but this is not the same as
advancing towards a Ph.D.

Don't equate a high-profile implementation project with high-impact
research.  Tonight I was reading the 8th RDS (reliable distributed
systems) proceedings, and I recall several "classics" [Lamport78]
[Gifford79] [Lampson81] [Reed83] that were never implemented by their
authors.

Disclaimer:  These are the views of an (?impartial?) former systems
person.