davecb@yunexus.UUCP (David Collier-Brown) (06/20/89)
"What Will We Do With All Those Cycles In The Year 2000?"
William A Wulf, NSF.
Keynote Address, Summer 1989 Usenix in Baltimore,
Given Wednesday, June 14th, 1989.
This is a short description of Wulfie's speech, with questions and
answers, done in point form from from my scribbled notes.
Responsibility for errors and omissions is entirely mine.
--dave
Description of the NSF
-- They're in their planning cycle for the next 5 years
In particular they're looking for technological drivers.
-- In their view:
1. Languages and operating systems will change in detail,
but they'll be named the same.
2. Bandwidth, on the other hand will go up enough to change
the way we use it.
-- in 1995-98, we'll have a 1 gigabit/sec national backbone,
with 200 institutions on it, slower area backbones with
300-odd institutions each and they will have 5 mb/sec t1
links to 2000 institutions each.
-- last year we had a 56 kb/sec backbone
This is a 100-fold increase, from slow to roughly memory speeds.
-- We are one of the few industries which lives with powers-of-two
increases, and are used to the fact that qualitative changes can come
out of quantitative ones.
-- So what will we use the power for?
1. Images
2. Decent front ends to systems/programs
The NSF would like to use some of it for the "Collaboratory".
Why fund this?
-- Its "our" way of trying out things, so lets build a prototype
and see where we get
-- Its large enough to be interesting, but still constrained
-- Its within the state of the art
-- Its a good idea & good public policy
its a multiplier of existing investment
we're losing scientists 6 engineers per capita over time
What is it, and what does it imply to CS?
-- The second most important laboratory tool is the blackboard
(the most important is the coffee pot (:-))
-- Teleconferencing is a failure
-- Remote instrumentation is currently useful
Examples:
Radio-telescope in Hawaii has its control
room in LA, but it should be usable from anywhere on
the net, not just one LA node.
100,000s of pc boards are done by labs every year,
because they can't get access to ASIC building-blocks from
where they are.
-- Digital libraries are on the horizon
(knowbots (:-))
-- Lab notebooks can be partially computerized
eg, mathematica as a pocket calculator
-- These same arguments can be extended to the commercial world.
eg, financial analyists.
The big noticeable changes:
-- The ratio of bandwidth to latency will change
-- Unix will change
User interface
VLDB
Scientific-data DBs
eg: global seismic db to receive & stores 1 terabit/day
Slow real-time
Questions from the floor:
1) sounds like the construction of the freeway system, but is it going
to be "free"? Ie, will it become available generally?
Its a hard question: first the research community, then broaden it to
the private sector, eventually available to anyone.
Its also not all under NSF control, nor is it profitable to industry
right now.
2) How do you get people to collaborate?
Just provide the opportunity. We really have to build the system
to be able to see: our current ones are too narrow-banded for real use
and attract frivolous ones.
3) What about computer-phobics?
Thats actually a funded research area. Rockefeller U held a conference
on the fundamentals of collaboration and propose an initiative. Psychological
effects of cold fusion claim recently was to cause widespread collaboration,
partially by the net.
4) Funding?
President and congress have promised to double NSF funding over 10 years.
5) The net is an example of an "active agent"..
6) How do you change human nature?
We can't.
7) Problems of scale: bandwidth alone is insufficient, organization
is lacking, coordination can be impossible.
Directory services (similar to white & yellow pages) are in process
for the 1991-95 timeframe.
8) The Lab notebook is both a scientific and legal tool. How do we
keep the lawyers from choking off scientific progress, as seems to
have happened already?
The solution is not available in copyright or patent. The National
Academy of Science & parallel organizations are trying to deal with this
9) Researchers are helped by this, what about science educators?
This net is both: its the "National Research & Education Network",
will spread at least as far as 3-year colleges. Below that state politics
interferes.
--
David Collier-Brown, | davecb@yunexus, ...!yunexus!davecb or
72 Abitibi Ave., | {toronto area...}lethe!dave
Willowdale, Ontario, | Joyce C-B:
CANADA. 223-8968 | He's so smart he's dumb.