\documentstyle[11pt]{report}
\begin{document}

\noindent{\bf Report: March 1, 1994}
\vspace{12pt}

\noindent Trip to NIST, March 1, 1994---this was conversations held
during closing sessions of workshop held to identify NII technology
areas.

I met with Dr. Shukri A. Wakid, Chief of the Advanced Systems
Division, and David Fisher, Program Manager for Software Proposals in
NIST's ATP Program.
\vspace{10pt}

\noindent{\bf i) Wakid}
\vspace{7pt}

\noindent NIST has technical side arranged into eight areas.  Wakid is
part of computer systems area head by Lyons.  This area has around 250
employees\@; Wakid heads up 90 people.  NIST only funds external groups
through Advanced Technology Program (ATP) and Manufacturing Extension
Program (MEP).  There is a companion Department of Commerce Agency,
NTIA, which funds network infrastructure.  (I think NYSERNET has
contacts here).

Wakid's division produces performance evaluation chips available on
Intel Paragon systems.  Wakid will give us boards with these chips on
them for either VME or SBus.  They are accessed by simple subroutine
calls from user code. They provide nonintrusive performance monitoring
in a time of $1.1\mu\,{\rm s}$.  They also have software to identify
performance bottlenecks in programs with an ingeneous technique
inserting wait periods.

He views NPAC as national leader in NII applications and has studied
efforts in Iowa (failure?) and North Carolina---more impressive.

NIST is on a Washington, D. C. dual 2.4~gigabit/sec ATM network
including ARPA, NSA, and so on.  He would like NPAC/NYNET to link with
this.

His division provides speech recognition, handwriting, and fingerprint
benchmarks for both training (neural networks) and evaluation of
methods.  His division provides benchmarks (TREK?) used by Liz Liddy
and others in ARPA Tipster program.  He is interested in more general
text benchmarks and was interested in askERIC and parallel oracle.

He voluntarily spend three hours with me and his interest in
collaboration is genuine.  He does not, however, have funding except
to give us NIST technology.
\vspace{10pt}

\noindent{\bf ii) Fisher}
\vspace{7pt}

\noindent The ATP program will grow in annual budget from \$10M in 1991
to \$750M in 1997.  The review procedure is very formal (so cannot be
influenced by program manager or congress) and the merit is 30\%
technical (but this is also a go/no go) and fully 70\% on business
case.  The goal is to produce useful products and business plan must
demonstrate this.  They expect around 1000 proposals this year, and
historically have funded 9\%.

All ATP must have technical lead in a FOR PROFIT corporation and all
intellectual rights must be held by such a corporation.  Universities
can have subcontracts, royalties, etc.

ATP's are either single corporations or joint ventures.
Organizations, such as MADIC can have administrative lead in joint
ventures.  Matching funds are required.

\begin{enumerate}
\item Federally derived funds are not allowed (e.g., I, R, and D
funds).
\item State funds are allowed.
\item Personnel salary and fringe benefits are allowed in kind
contributions.
\item Some small amount of equipment cost sharing allowed.
\item ATP will assume, without any justification needed, that indirect
costs for single corporation projects are equal to direct costs.  This
allows a company to propose that ATP funds personnel and company
contributes indirect costs.  For a small business, where indirect cost
is less than direct cost, this can be a good deal.
\end{enumerate}

In a traditional ATP, they prefer that matching comes from
organizations that will use/benefit from technology, for example,
computer vendor (e.g., IBM) or user organization (e.g., Niagara
Mohawk) could match federal funds for a proposal to develop power
industry software.  We could, less plausibly, use state funds.
Niagara Mohawk or IBM would be better as showed they cared about
software (i.e., gave more credence to business case).

Single corporation ATP's have a three year, and joint ventures a five
year time limit.

The new base ATP program will be announced in a week.  You can send a
white paper for evaluation---they reject 80\% of these.

There will be a new type of ATP announced in April.  This is to address
emerging areas where the ``industry'' does not yet exist.  You are
welcome to suggest projects for focus areas at any rate.  Currently,
Fisher has 95 white papers in software area.  He has divided into
three broad areas.
\begin{enumerate}
\item software to help software companies become more productive.  One
reason for importance of this is trend to contract software to India
and Eastern Europe, which have much cheaper personnel costs.
\item NII infrastructure software (e.g., collaboration technology).
Enterprise Integration is one area here and NIST will sponsor a
workshop in this at the end of March.  The software industry is very
proactive here.
\item Educational and Training Software.
\end{enumerate}
1) and 2) are reasonably clear and could be focus areas in April
announcement (NIST is making decisions as to focus areas now).  3) is
very interesting, but currently nearly all white papers are from
potential customers and only a few from potential suppliers.  NIST
intends a set of workshops to establish partnerships and issues over
the next year.  3) could be a later focus area.

White papers from training customers include the Teamsters (``flight
simulators'' to train truck drivers---this seems to me an interesting
opportunity for a home PC/videogame product).  Another came from
cement producers as current wonderful cement is ruined by mistakes by
those who pour it incorrectly.
\newpage

\noindent{\bf Minutes: March 1 meeting with Paul Hunter}
\vspace{12pt}

Paul Hunter is Program Manager at NASA headquarters for their new NRA
on ``Public Use of NASA Databases Over the Internet.''  The formal
full announcement and public briefing on this are due in next few
weeks.  He expects some 300 proposals and considers collaboration with
JPL as reasonable, even though Goddard has lead in Earth databases.

The InfoMall proposal could combine
\begin{enumerate}
\item education (Living Text Book)
\item entertainment (AGE)
\item city planning
\item defense dual use
\end{enumerate}

We need to contract Mappower regarding 3),  Rome and Ultra regarding
4).  We need an NPAC lead person here.
\begin{center}
{\bf Notes on Testbed}
\end{center}

\noindent{\bf i) Original Discussion}
\vspace{7pt}

\noindent We see that many small (10--100 nodes) general and
application specific testbeds are feasible and useful.  It is not
practical to build a dedicated ``one million node'' testbed.  Rather,
the NII is the testbed.  Spector emphasized that NII commercial
products must be tested in the real world of the NII.  Only this had
the bizaare mix of requirements that typified real customer needs.  If
you built a separate (large) testbed, it would become used in
production anyway and so, in fact, become part of the NII.

We need to arrange that the NII has enough protection to support test
and development function.  We need to
\begin{description}
\item[\rm a)]Protect NII itself from ``thrashings'' and other
unintended misuse---here we probably need even more protection for
malicious users than for those that produce accidental damage while
testing.
\item[\rm b)]Protect those testing from ``spending too much money'' with a
faculty program.  Testbeds are needed not just to provide an
environment that can be thrashed, but also to provide unique resources
not available on the NII.  There was agreement that one needs to allow
small businesses (the virtual garage) to access suitable testbeds
(maybe just the NII itself).  It is possible that providing NII test
and evaluation environment would be a viable business (analogous to
``learn-to-drive'' companies and off road and speedway facilities).
\item[\rm c)]We discussed and ``rejected'' certifying users (driving
licenses) of digital highway.
\item[\rm d)]Use of simulation to produce a virtual million node testbed.

c) and d) have some limited value.
\end{description}
\vspace{10pt}

\noindent{\bf ii) Application-Specific Testbeds}
\vspace{7pt}

\noindent We discussed the need for application-specific
testbeds that would play the same role for the NII that the
grand challenges do for HPCC.

Particular focus areas to be tackled by a dedicated testbed
include
\begin{itemize}
\item national challenges, such as health care and financial
services;
\item particular technologies, such as human computer
interface, massive I/O, and real time and other
synchronizations; and
\item performance scaling with many nodes.
\end{itemize}

In HPCC, we distinguish the application development group
(this is a mix of application scientists and computer
scientists) and resource providers.  The latter are NSF,
NASA, DOE, and so on---``supercomputer'' centers.  There
will be one important change in NII testbeds as such
testbeds are likely to be geographically distributed and so
more like the generalization of the ``NSF Metacenter'' and
gigabit testbeds rather than individual localized centers.
However, one is still ikely to need dedicated ``resource
providers.''  An experimental geographically distributed
testbed will be as hard to run as a localized
(experimental/dedicated) supercomputer center.

If the federal government funded national challenge testbeds
(e.g., NIH, a health care testbed; EPA, an environmental testbed;
NSF, a technology testbed; ARPA, a global grid testbed;
Treasury, a financial services testbed; NASA, a
manufacturing testbed; DOE, an energy management testbed),
one could first link them together and then integrate into NII itself.
\newpage

\noindent{\bf Exemplar Applications}
\vspace{7pt}

\noindent This working group proposes a software model where
NII capabilities are built up from one or more coarse grain
objects or services.  Here, we examine three representative
NII applications to illustrate and verify this approach.
These applications are quite rich and illustrate several
possible services.  However, our choice does not imply that
these are the three most important applications, and
further, they do not by any means cover all possible
services.  We recommend that detailed studies be conducted
to refine (deep examination of each application) and broaden
(many more applications) our illustrative discussion.  We
chose two applications from the workshop briefings---namely,
financial and health care services.  Our third exemplar
application is rapid prototyping and distributed research
and development---chosen because it can require more
flexible and less robust services.

\begin{tabular}{l|l}
\multicolumn{1}{c|}{\bf Generic NII Objects}&
\multicolumn{1}{c}{\bf Realization in Financial Services} \\
\\
User interface (client)&Easy to use customer interface \\
User interface (service objects)&Easy to use banker
interface \\
\\
Security monitor&Test security \\
Security model&model and forecast vulnerability \\
\\
Real time synchronization object&Real time processing of
financial \\
Distributed database object&records with authorization \\
Authorization object& \\
\\
Encryption/descryption object&Make secure and examine documents, \\
Authentication object&certificates, receipts, signatures \\
Tamper/copy proofing and & \\
testing object \\
\\
Digital signature object&Identify customers and bankers \\
Biometrics service& \\
\\
Real time simple anomaly &Examine single and
multiple \\
detection object&transactions to detect possible fraud \\
Batch complex anomaly &and errors \\
detection object& \\
\\
Simulation object&Compute financial models and \\
Visualization object&visualize \\
\\
Image compression object with&Transmit and store images of \\
size/fidelity tradeoffs&financial paper records \\
\\
Contract object&Implement legal computer contract \\
\\
Digital cash object&Implement electronic checks, credit, \\
&and bartering 
\end{tabular}
\newpage

\noindent{\bf To: Barbara Mihalas}
\vspace{10pt}

\noindent Possible responses to curriculum development for
undergraduates' solicitation
\begin{enumerate}
\item Science for Nonscientists

Development set of NII (Mosaic) modules to use in course of
modern science for nonscientists.  This will start with
computer science modules to develop understanding of
technology used to display further modules.  Modules will be
designed to exploit virtual reality headsets if available.
Science modules covering highlights of modern physics,
chemistry, and biology will be produced.  As available,
video, simulation images, and text on demand will be
incorporated into each module.  This is based on existing
Syracuse course successfully taught last semester by Physics
department.  Mosaic was successfully used in modules on
computing and neural networks.  Here, we will extend the use
of NII technology to the entire course.

\item Computational Science for the NII

We have developed graduate and undergraduate courses in
computational course aimed at three audiences.
\begin{itemize}
\item Computer Science and Engineering
\item Applied Mathematics
\item Science and Engineering application fields using
simulation
\end{itemize}

The courses are focussed on base hardware, software, and
algorithmic technologies.  Examples are parallel computers,
data parallel Fortran, and iterative solution of partial
differential equations.

We will develop an undergraduate course focussed not on
simulation but on the parts of computational science needed
in information (NII) applications.  Now the audience is
\begin{itemize}
\item Computer Science and Engineering,
\item Applied Mathematics,
\item Students in IST---these are future developers of
digital libraries and corporate information systems,
\item Students in Newhouse---the future multimedia
journalists,
\item Students in Maxwell---future users of multimedia to
aid political decision making.
\end{itemize}

Now base technologies would be illustrated by parallel I/O
subsystems, databases, and image compression in hardware,
software, and algorithm areas.

As in current computational science courses, the curriculum
has two synergistic thrusts.  It will equip computer
science students to develop technologies to support the NII.
It will equip application types to use the NII as they
understand the innovative technology and how best to use it
now and in the future.

This course will be developed as a set of linked (Mosaic)
NII objects---one object per technology.
\end{enumerate}

\end{document}