rick@cs.arizona.edu (Rick Schlichting) (06/11/91)
Richard D. Schlichting (rick@cs.arizona.edu)
Associate Professor
Department of Computer Science
The University of Arizona
Tucson, AZ 86721, USA
[This report represents the personal opinion of the author, who was on
sabbatical in Japan from Dec. 1989 through July 1990. The sabbatical
was supported in part by grant INT-8910818 from the NSF U.S.-Japan
Cooperative Science Program.]
---------------------------------------------------------------------------
On Thursday morning, I was met at my hotel by my host, Prof. Eichii
Miyamoto from Hokkaido University. Since the hotel was very close
to campus, we walked to his office, which was gave me a nice chance
to look at the university and get some general impressions. Overall,
I would say that the feeling is one of spaciousness and newness, making
it appear much more like an American university than many of the other
Japanese universities I visited. This stems partially from the fact that an
American was instrumental in setting up the university in the early
part of this century and partially because Sapporo itself is much
newer and more spacious than other Japanese cities. The university
seems very comprehensive, as I saw everything from an agriculture
school to a medical and dental college.
Our first stop was the main College of Engineering building, where
we stopped to get an English-language brochure on the College. We then
walked up and over a covered walkway to the Information Engineering
Building, a 1 and a half year old building where Miyamoto's laboratory
is located. Once there, we sat and chatted for awhile about the
general setup of the lab. A few points that he made:
-- At the current time there are approximately 7 undergraduates,
7 MS students and 3 PhD candidates associated with his lab. He
indicated that this number was about average for the other labs
in the department as well.
-- They have been connected via email to other Japanese universities
for about 3 years. This is is used extensively to keep in touch
with colleagues elsewhere in Japan, and is especially important
given Sapporo's relatively long distances from other cities.
-- Japanese-language interfaces are no longer the stumbling block
that they were just a few years ago.
I ended up giving him the rundown of my university and department as well,
and we both found the differences between the two educational systems
interesting.
After about 45 minutes or so, we left to go talk with his students
and the other professors associated with his lab. They each introduced
themselves and briefly characterized their research area; most were
working in object-oriented systems, with some AI as well. Two students
then got up and took turns describing their work on a system they called
Kamui. Kamui, it turns out, is an object-oriented system in which
objects can be written in different programming languages. The interface
of each object is defined using a protocol that described the operations
the object exports and the types of the arguments. Another object wishing
to use these exported operations name the same protocol. A single object
can also have multiple protocols describing different interfaces; although
similar to inheritance, it differs in that it need not form a hierarchy.
Kamui also has a shell, which is apparently used to make Kamui objects
out of non-Kamui programs. This shell also interacts with some user
interface work they have done to allow, for example, the mouse on
one machine to control the cursor in the window of another. Differences
in value representations between different languages is handled by
translating into the target representation before sending.
This sounded very related to some of my own work, so I then got up
and gave an impromtu overview of our MLP system. They had not heard
of the system previously, but were familar with better know related
systems such as the HCS work at Washington, Mercury at MIT, etc. In
general, they seemed very much aware of the issues involved and had,
in my opinion, constructed in Kamui a very interesting solution.
Another student then got up and described his work on a hierarchical
nameserver scheme for wide-area networks. The main selling point
appears to be that requests for application programs that happen to be
on failed nodes are automatically rerouted through the nameserver to another
copy. The focus of the scheme is primarily on finding relatively
static objects such as code, so that they have not had to address
the problem of keeping the copies of the objects consistent in the
face of failures.
At that point, the meeting broke for lunch. Afterwards, I gave
a couple of formal research presentations, and was then treated to a nice
tour of Sapporo by my host.
----------------------------------------------------------------------
Prof. Eiichi Miyamoto
Dept. of Information Engineering
Faculty of Engineering
Hokkaido University
Sapporo 060, Japan