Kruskal MH Sent Mon Nov EST

Found at: ftp.icm.edu.pl:70/packages/netlib/research/permission

From research!kruskal Mon Nov 26 16:30:03 EST 1990
	Kruskal  201-582-3853 MH 2C-281  Sent Mon Nov 26 16:30:02 EST 1990

To: Jon Bentley
Cc: Eric Grosse

Thanks for re-sending the information from David Johnson about "l2fit".

At the same time, my interest is and was not particularly in l2fit, but
Also, my interest is and was not only in my obtainng access to the
s why I suggested to you that you "ship" them to the suite of
Tenth Edition Unix systems around here.

As an alternative, how about entering them into the research!netlib
collection of source programs?  I hope that Eric Grosse would be happy
to add them to that data base.  That would make them all generally
available, and would not require the same degree of consent that


------m jlb kruskal Sat Dec 1 23:46:04 EST 1990
To-: jlb kruskal
Subject: one-liners in netlib

Naturally, I'd be happy to distribute these by netlib.  There are a
few new aspects to think about.  For example, when someone asks for
l2fit, should they be sent just the awk script, or also the Port
the pain of installing port and awk?  Is it easy to describe exactly


From nls Wed Dec  5 06:57 EST 1990
To: Joe Kruskal
cc: ehg and jlb

but the only tool that has stood up well to use has been Bentley's "
l2fit". The reason is that all the rest (ode and pde solvers, etc)
manipulation of the output, or even further solves. That is,
only l2fit seems to be a nice "self-contianed" application,
the rest require an "open" language (rather like ratfor) so that the
user can embed the solution in his larger problem.

n would not be a good idea: we are replacing (slowly) them with a
full-blown open language. That second pass should be worth putting
nto netlib when its ready.

					Norm Schryer

Eric: I think we can come up with a README file that isolates the awk
	dependencies (maybe even the V10 src for it), make a l2fit.f
	file with the port3 src tree for nl2sol, etc, to make l2fit
	a reasonable self-contained package, so that folks can say:

		send src from l2fit

	and then simply "sh <mail_bundle; make" to have a working l2fit.
	I'd be glad to make (help make) that happen. Jon you too??

From research!kruskal Mon Dec  3 10:07:42 EST 1990
	Kruskal  201-582-3853 MH 2C-281  Sent Mon Dec  3 09:55:33 EST 1990

To: Grosse, Bentley
Subject: Bentley's one-liners in netlib

Here are my suggested answers to most of Eric's questions.

(1) Requesters for l2fit should receive only the awk script, plus
users might find installing these things difficult.

(2) It is reasonable to imagine experienced Unix users installing the
of his system administrator to act as requester, and surely the
administrator would not find installation a problem.

(3) Re flavors of awk which will suffice:  whatever information can be
user should simply be cautioned that not all awk's will work.


From research!kruskal Wed Dec  5 13:10:25 EST 1990
	Kruskal  201-582-3853 MH 2C-281  Sent Wed Dec  5 12:45:35 EST 1990

To: schryer
Cc: bentley fernandez grosse kernighan
Subject: Putting l2fit and its relatives in netlib

l2fit and its relatives available through netlib, and that you
are giving l2fit the final bit of polish necessary for such distribution.

You indicate that the relatives need further development before
they are ready for general distribution.  I assume this viewpoint
s the concensus of the authors, and I look forward to seeing how
much better these tools become.


From jlb Wed Dec  5 13:59:29 EST 1990
l2fit gabfest when I return (late Friday night).  Jon

From jlb Wed Dec  5 14:02:43 EST 1990
Before anybody does any work, I believe that I have an l2fit almost ready
to go in /n/coma/usr/jlb/l2fit, complete with README.  I'll look into it

From jlb Tue Dec 11 16:43:22 EST 1990
/usr/jlb/l2fit (on coma) contains the l2fit stuff.  I ship it by
	make.bundle | mail victim.name
How to put it in netlib?

From jlb Tue Dec 11 16:46:30 EST 1990
/usr/jlb/cputimes/time.c contains a (fairly undocumented) C program.

------m jlb Sun Dec 30 09:50:51 EST 1990
For cputimes/time.c, there needs to be some better explanation of how to