**Today's Topics:**

- New, improved NA-NET
- NA-NET statistics
- Change of Address for Jim Pool
- Special Issue, ACM Transactions on Graphics
- PDE solver needed
- Format bug in PARANOIA
- Triangularization of 2-d Domains
- Query about singular forcing functions

From: Mark Kent <kent@na-net.stanford.edu>

Date: Fri, 19 Feb 88 20:55:11 PST

Status: RO

Quite a while ago I moved the na.<lastname> facility from

navajo.stanford.edu to another machine (almost transparently).

Now we can bypass score.stanford.edu too.

Following are the new instructions for the NA-NET.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% Requests, comments or problems to nanet@na-net.stanford.edu

% Submissions for broadcast to na@na-net.stanford.edu

%

% Note: Most people will be able to mail to host "na-net.stanford.edu".

% However, for those who can't, the mail can be routed through another

% machine: e.g., na%na-net@score.stanford.edu, or

% nanet%na-net@score.stanford.edu

%

% The only people who may have problems getting to na-net are those

% on the Arpa-Internet who are using static host tables rather than

% name resolution software. I believe that all the major gateways

% to the other nets have good software and there should be no problem

% with mailing from bitnet (netnorth, earn) or csnet sites. All

% uucp sites that enter the Arpa-Internet through a "good" host

% will also have no problems e.g., rutgers.edu, hplabs.hp.com,

% uunet.uu.net. This list is not exhaustive.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

The "old" way will still work for quite a while

(i.e., na@score.stanford.edu and na.<lastname>@score.stanford.edu),

but I think that score itself will disappear this year.

Coming features: SOON, but not YET, we will drop the "na." in

the na.<lastname> facility.

-- Mark Kent

------------------------

From: nanet@Bravery.stanford.edu (The NA-NET)

Date: Fri, 19 Feb 88 21:55:34 PST

The digest is currently mailed out to 821 distinct addresses.

The mailing list is split over eight cpu's roughly according to

(what I perceive is) the distance from Stanford of the recipient.

Here is the current breakdown:

#addresses #connections "sample distance"

group1 90 28 trans-atlantic Internet

group2 83 40 East coast USA

group3 100 37 .gov and .com hosts

group4 107 23 Midwest USA and uunet

group5 61 39 Washington, Utah and Texas

group6 72 35 California

group7 236 1 bitnet (through Stanford gate)

group8 72 1 csnet

The number of connections is the minimum number of mail connections

that must be made by our hosts in order to get the mail off our systems.

(guess which one is my workstation.)

Here are the usage figures of the na.<lastname> facility

since we started keeping track.

6730 messages from Feb 12 to Aug 19 (1987)

3133 messages from Aug 19 to Nov 8 (1987)

1966 messages from Nov 8 to Dec 14 (1987)

1405 messages from Dec 14 to Jan 19 (1987-88)

1856 messages from Jan 19 to Feb 19 (1988)

So, in a year and a week, we have processed over 15,000 messages from

one individual to another. The na.<lastname> facility has

over 1050 entries. You can get the current list by mailing

to netlib@anl-mcs.arpa or netlib@research.att.com with the

message

send nalist from misc

--Mark Kent

------------------------

From: James Pool <pool@lbl-csam.arpa>

Date: Sat, 13 Feb 88 10:29:52 PST

Effective February 9, I resigned my positions as Executive Vice-President

and as a Director of The Numerical Algorithms Group (U.S.A.) Incorporated.

I plan to take a respite and then seek a new position.

During this period, please use the following temporary points of contact:

US Postal Address

James C. T. Pool

9 South 532 Drover Lane

Downers Grove, IL 60516

Telephone

(312) 985-3305

pool@anl-mcs.arpa

------------------------

From: Richard Bartels <rhbartels%watcgl.waterloo.edu@relay.cs.net>

Date: Sat, 13 Feb 88 15:02:05 EST

ACM's Transactions on Graphics solicits papers for a special

issue on computer-aided geometric design and geometric modeling.

Topics of interest include, but are not limited to:

* parametric curves and surfaces

* algebraic curves and surfaces

* procedurally defined curves and surfaces

* solid modeling

* polynomial and spline techniques

* blending and filleting

* geometric continuity

* sculpting paradigms

* engineering applications

* interaction techniques

* curve and surface fitting

Five copies of each submitted manuscript should

be sent to the following guest editors:

Richard Bartels and Ron Goldman

Computer Science Department

University of Waterloo

Waterloo, Ontario N2L 3G1

The schedule is as follows:

Submission deadline 15 September, 1988

Notification 15 January, 1989

Revision 15 March, 1989

Publication July, 1989

For style requirements, please see the July, 1987,

issue of ACM's Transactions on Computer Graphics.

Prospective authors may discuss their plans informally

with one of the guest editors by contacting them at

(519) 888-4534 or via their electronic mail addresses

rhbartels@waterloo.edu or na.rbartels@na-net.stanford.edu

rngoldman@waterloo.edu

------------------------

From: Arunava Majumdar <majumdar@ucbvax.Berkeley.EDU>

Date: 18 Feb 88 00:21:12 GMT

Organization: University of California, Berkeley

I need a package to solve the following form of

a PDE:

u [n] + u [n] = f(u[n])

x y

where u is a n x 1 vector. To solve this system of

equations, 2n boundary conditions are needed. In my

problem the boundary values of u are distributed

over all the boundaries, ie, x=0, x=l, y=0, y=L.

I need to find the field of u within these boundaries.

Is there any package which solves this type of system

of PDEs? If so could someone let me know the name and

if possible how to get hold of such a package.

Thanks.

--- Arun

------------------------

From: Jim Kirkpatrick <ZSYJKAA%WYOCDC1.BITNET@CUNYVM.CUNY.EDU>

Date: Wed, 17 Feb 88 16:37:44 MST

Sorry if this duplicates earlier postings... I recently got a copy of

PARANOIA and tried running it on our CDC Cyber 840 and discovered a minor

bug. In the single-precision Fortran version there is a subroutine PARTUF

which has a FORMAT statement with a label of 4815. In this format there is

a literal character string 'zname' which needs to be changed to an A2

specification or you might end up trying to print a character string via

an E16.8 specification. I do not know about the double-precision version.

I have dropped a note to Richard Karpinski about this.

Has anybody found any other bugs?

------------------------

From: Cleve Ashcraft <ashcraft@yale-zoo.arpa>

Date: 20 Feb 88 17:51:50 GMT

Organization: Yale University, New Haven, CT

I am interested in ordering sparse matrices arising from 2-d finite

element problems which use linear triangular elements. The density

of the grid points in the domain should be roughly equal throughout

the domain, and the triangularization should be "good". The

triangularization should exhibit no apparent regularity, this last

property is very important.

Can anyone recommend book or journal articles on this subject?

Thanks in advance.

Cleve Ashcraft

C. S. Dept

Yale University

ashcraft@yale.arpa

203 -- 432 -- 1221

------------------------

From: Dick Beyer <dbeyer@gauss.am.washington.edu>

Date: Sat, 20 Feb 88 12:56:21 PST

I am interested in numerical methods and error analysis

for PDEs with singular forcing functions. For example,

a delta function source whose location and strength depends

on the solution of the PDE.

In particular, I would like to use a uniform finite

difference grid. Since the source location will not, in

general, be at a grid point, some interpolation will be

necessary. I am interested in deriving or locating second

order accurate methods for such problems.

I have not had any success searching the literature and

wondered if anyone might point me to a possible reference or

author.

Please address email to

dbeyer%riemann.am@cs.washington.edu

Thanks

Dick Beyer