From: MIT component of CS-TR project
(Jerry Saltzer and Greg Anderson)
To: Bob Kahn, CNRI
Subject: Quarterly Progress report, October 1--December 31, 1993
C. Talks and papers
A. Mostly "administrative" things, relating to startup of on-line
technical report services:
We installed processor upgrades to two IBM RS/6000 servers used for
indexing services and for program development. Our server complement
now consists of two model 530E's and one model 530, each with 128
MBytes of RAM. All were expected to have been upgraded to contain 0.5
GByte of RAM by now. As mentioned in the previous quarterly report,
IBM's internal restructuring has severely compromised its ability to
deliver on promised grants, and they have fallen far behind in
deliveries. In addition, grant bookkeeping problems within IBM have
roadblocked our placing a final round of hardware orders, for
development workstations, so our students are using borrowed, creaky,
and obsolete facilities. We have requested help from sources high
inside the corporation and are hopeful that we will see some progress
in this area in the coming quarter.
Mitchell Charity established FTP and limited HTTP servers for our
materials, for use by the other project participants. The FTP site is
found at cstr-ftp.lcs.mit.edu while the HTTP site is at
cstr-http.lcs.mit.edu . He also set up an experimental domain name
service using the name ltt-ns.lcs.mit.edu .
Considerable effort was expended tracking down technical information
about the hardware interface strategies available for the Fujitsu line
of production quality scanners. We established from this work that
this scanner appears to be the best bet for the kind of workload we
are expecting over the next year or so. We also established that
scanner management software is still in much flakier condition (this
situation is well known, but we had hoped maybe it had changed) so we
worked through a strategy involving purchase of a scanning station (A
Macintosh Quadra 840AV with 4 Gbytes of disk and 64 MBytes of RAM), a
simple scanner (HP IICx) and trying out several software packages with
that hardware complement.
We established that the market structure in this area is such that the
cheapest way to acquire this kind of system is by acquiring major
components (scanner, cpu, disks, monitor) from different vendors and
assembling it ourselves, rather than by single-vendor purchase of a
complete system. Although this strategy will often save money, the
unusual aspect of this market is that all the equipment is
customer-setup, which eliminates the usual system-integration benefit
of working with a single vendor. Accordingly, we got on the phone,
obtained a large collection of quotes, and placed orders. The last of
the hardware arrived during the last week of the reporting period, and
assembly of the scanning system is scheduled for early January.
In a similar move, Mitchell Charity spent a fair amount of time
exploring the current market in magnetic disks and disk packaging.
That exploration led to two conclusions:
- the discounts available for purchasing magnetic disks in large
quantities (on the order of 10%) are small relative to the rate at
which market prices are dropping with time (5% per month) so there is
no point in buying ahead of need. A just-in-time ordering system
(perhaps having the computer system note that it is filling up the
last disk and automatically printing a purchase order for the next
one) seems to be the right strategy.
- packaging at the level of placing 3.5-inch magnetic disk drives
in arrays has not kept pace with the disk drives themselves. The best
available disk packaging systems are very space-inefficent, wasting
75% of the box volume. He did find one very promising example;
Digital Equipment has developed a nicely space-efficient system of
disk packaging, but DEC hasn't yet figured out that it should market
this packaging system as a separate product. (Recent indications are
that this marketing problem may soon get fixed.)
On this basis, we ordered an interim packaging system and a modest
number of disks to get us started with some space to store images that
will soon be coming from the scanning system. We are hopeful that our
next order of a packaging system may be able to take advantage of the
developments at DEC.
To obtain an additional storage server for replication experiments we
worked out a deal with M.I.T. Information Services in which they will
provide us with a currently disused RS/6000 520, we will equip it with
magnetic disks, and we will share the resulting server between the
CS-TR project and the TULIP project, another image-delivery experiment
in which we are jointly participating. In addition to the opportunity
to share hardware resources, we hope to get some additional benefit by
mixing and matching software packages, some of which we are developing
and some of which they are developing.
Since the contract with CNRI was signed at the end of the previous
quarter, making funding commitments possible, it was finally possible
for us to advertise for thesis students to join our research group.
The advertising proved effective, and we now have two new graduate
research assistants, one student doing an advanced undergraduate
project (our department's new name for the senior thesis) and three
new undergraduates, in addition to the two undergraduates who were
working with us at the beginning of the quarter. The new members
joined the group too late for any progress to be reported this
quarter, but they should contribute to a substantially larger report
of work started and work-in-progress in the coming quarter.
One of these students, Ali Alavi, did begin to investigate the use of
the Domain Name System as a name resolver to convert from report
identifers found in bibliographic records to the resource locators
needed to retrieve a document.
On the basis that most of our discussion in recent meetings has been
focused on technical issues, Greg Anderson has been working with
Rebecca Lasher of Stanford to develop a list of issues that this
project raises from the perspective of the librarian. It is their
intent to present this list at the forthcoming meeting at Stanford in
February, and thereby to initiate discussion.
Mitchell Charity created some tools to print CS-TR reports from
images. This experimental package, when given the name and page
numbers of a Berkeley TR, will locate that TR at the Berkeley server,
grab the page images, and send them to a printer here at M.I.T. We
suspect that this is one of the few examples of interoperability
currently extant in the project.
Mitchell Charity also began work on fast image scaling and X display.
Expect to hear more on this topic next quarter.
C. Talks and Papers
Although we do not normally report talks given internal to M.I.T., one
this quarter is of possible wider interest. On November 16, the
M.I.T. Academic Council (A group of senior administrators and deans
that have day-to-day responsibility for running M.I.T) devoted one of
its regular meetings to a discussion of the future of scholarly
communication at M.I.T. Jerry Saltzer presented to this group a
vision of how technology is likely to affect the library of the
future. This meeting is an indication that the topic is beginning to
rise high on many agendas.
Greg Anderson, Mitchell Charity, and Jerry Saltzer attended a 1.5 day
meeting of the CS-TR project participants at CNRI headquarters in
Reston, Virginia, on November 25 and 26.
Mitchell Charity, Marilyn McMillan, and Jerry Saltzer participated in a
two-hour PictureTel conference with the other CS-TR project
participants on December 2.
Return to Library 2000 home page.