3.333 workstations; IT; software development (174)

Willard McCarty (MCCARTY@VM.EPAS.UTORONTO.CA)
Mon, 7 Aug 89 22:31:14 EDT


Humanist Discussion Group, Vol. 3, No. 333. Monday, 7 Aug 1989.


(1) Date: Sat, 05 Aug 89 16:58:37 EDT (105 lines)
From: "James H. Coombs" <JAZBO@BROWNVM>
Subject: MIPS; MACII screens; hypertext development

(2) Date: Monday, 7 August 1989 1904-EST (31 lines)
From: JACKA@PENNDRLS
Subject: IT; MIPSy workstations

(3) Date: Mon, 7 Aug 89 16:44:12 PDT (13 lines)
From: Paul Delany <USERAARY@SFU.BITNET>
Subject: MIPSy workstations

(1) --------------------------------------------------------------------
Date: Sat, 05 Aug 89 16:58:37 EDT
From: "James H. Coombs" <JAZBO@BROWNVM>
Subject: MIPS; MACII screens; hypertext development

A few quick comments from skimming/scanning discussion of MIPS, etc.

1. I use SUNs for database construction and servers. I find the SUN 3 about 4
times as fast as a MACII for database construction. Perhaps 2 times as fast
for serving. Others might have different experience. I have to allow at
least 24 hours to process the American Heritage Dictionary and build a new
database (27 meg.); lot's of extra processing but the database is highly
optimized.

2. I have not seen any software that is designed to serve scholarly research
or education at high levels. Sun Tools is nice. et++ is interesting. But no
special tools to help me in my research. It might be there, but I haven't
seen it.

3. We use SUN 3s and NFS to mount large filesystems on our MACs for program
development. High speed. Lots of disk space. Share files. We use a MACII
and NFS for our Intermedia database and documents. Performance is fine.

4. I am using a large b/w monitor on a Mac IIcx. I guess it's 21 inches diag.
At any rate, I almost never feel a need to open a window that stretches all of
the way from top to bottom. I find that the window borders help me keep my
place in the document. That means that I have free space on the "desktop" not
because I am trying to keep it free but because I don't need it all of the
time. Intermedia runs on this monitor without requiring modifications. (We
did have to change some dialogs recently to fit on the tiny SE 30 display.)
I don't know how much the monitor costs, but it is good.

5. I'm not sure how this one came up, but there has been some discussion of
the costs of developing a hypermedia system. Good software engineering is
very expensive. Maintenance is even more expensive than creation. There are
no easy answers. Some people say that there are NO answers, only tradeoffs
(D.A. Norman on cognitive engineering; G. Weinberg on systems design; etc.).

We had problems with Ingres. I had just developed a dictionary database using
CTree, a b+tree file management system. Victor Riley and I developed a basic
communications protocol. We implemented a client layer to slide under
Intermedia and a server with a pluggable database layer. Also the CTree
implementation of the database layer. This took two weeks to design and
implement. It took another couple of months to debug and get it up to product
quality. We had only one lingering problem: documents were left locked when
Intermedia crashed. Part of the solution was to debug Intermedia; the other
part was to use a "keepalive" option on sockets (and this required hacking
the kernel when A/UX 1.1 came out).

Ok, so we had a usable database and server in two weeks, with major bugs out
in a total of one month. We had to develop the communications protocol, but
we were given a hypermedia model, which we modified only slightly. We also
had a server building block already developed and in use for about six months.
Currently, we have about 8000 lines of code for the server, client, and
database layers.

Note that this does not include any interface, editors, or anything of the
sort. It also does not include the maintenance of lists of blocks and links
on the Intermedia client side. On the other hand, we were working only
about 1/4 time on this project once we finished the initial coding.

I will leave it to others to estimate what it takes to create a minimal
hypertext environment. My experience is that developing a good interface
takes a long time. It cost me about two weeks to put up a dialog for
spelling correction and get it right. 5 buttons, 1 scrolling tile, 1
edit text box, 1 static text box.

I currently take my best estimate and multiply it by 4. That just about fits
the two weeks it took to implement the hypertext server and the two months
that it took before we could forget about it. This is based on personal
experience. I don't know what formula others apply, but this one works for me.

My advice to humanists who are thinking about contracting for some software
work? Don't. I suspect that only 1 in 10 ever get something they can use.
Get your university to establish a staff of developers. That staff can
include students, who will get to work with experienced professionals, both
for their own benefit and for the benefit of the project. Most universities
will say that they cannot afford such a staff. If not, then how can
individual faculty or departments afford it?

Principles:

a. Software requires maintenance.
b. Maintenance is more time consuming than initial development.
c. Inexperienced developers need close supervision. So do experienced
developers (I have seen 40,000 lines of code developed over 2 years
thrown away for poor quality). I guess I should add that only
experienced developers have the qualification to provide this supervision.
d. This is not a problem for scholars and humanists uniquely. I'm beginning to
accept that software engineers should be licensed.
e. When negotiating for software, don't forget to include the costs of your
own time (and probable stress).


I guess this is enough for a start.

I acknowledge that there are some success stories: my guess is 1 in 10.

--Jim

Dr. James H. Coombs
Senior Software Engineer, Research
Institute for Research in Information and Scholarship (IRIS)
Brown University, Box 1946
Providence, RI 02912
jazbo@brownvm.bitnet
(2) --------------------------------------------------------------34----
Date: Monday, 7 August 1989 1904-EST
From: JACKA@PENNDRLS
Subject: IT; MIPSy workstations

Two quick notes:

1. ON IT: As organizations have a tendency to do, we in
Computer Services at the University of Pennsylvania had to
spend each Wednesday of last semester together writing a mission
statement for Computer Services. It is perhaps interesting
to note that those colleagues who came from administrative computing
spoke of computing as IT. Those of us from academic computing
didn't know what they were talking about even when IT was defined;
however, we embraced the term fully when we saw that IT could
mean most anything and would, if included in our mission statement,
allow us to take under our wing the library, the media center,
the typing pool and almost any piece of technology we could label
as technology.

2. The whole question about which machine is faster and more
powerful reminds me of my days in high school when we used to
run cars at night down the straight ways on Chapel St. in New Haven,
Connecticut. So what? I think it is relevant to point out that
software is the primary requirement along with personal choice.
Also, I liked to add that there still exists such machines as
mainframes which are certainly more powerful than any of the machines
mentioned yet in this discussion. The problem here is the
software and interface to a certain extent. Mainframes remain
fossilized in the ideas of the past unlike the SUN or NeXT which
are present tense concept, but certainly not futuristic enough
for my tastes.
(3) --------------------------------------------------------------17----
Date: Mon, 7 Aug 89 16:44:12 PDT
From: Paul Delany <USERAARY@SFU.BITNET>
Subject: MIPSy workstations

Thanks for the many useful comments on this issue, & may the
discussion continue. HUMANIST is an invaluable resource for
such decisions.
We have decided to go ahead with our plans for a Macintosh Lab,
helped by Apple's recent offering to us of Mac IIx's for $2900Cdn
(that's just the box, but still . . . ). Meanwhile, there's a chance
we can get free use of a Sun 3/50, declared surplus by another department,
to experiment with. They can also be upgraded to a 3/80 for about $1200.
Again, my thanks for the advice. Paul Delany, Simon Fraser Univ.