NSWC Mathematics Subroutine Library

[andreasnoack]

University of Alberta, http://www.ualberta.ca/CNS/RESEARCH/Software/NumericalNSWC/site.html hosts the mathematical functions from Naval Surface Warfare Center. Besides the cool name, it contains some functions not covered by the present library, e.g. the incomplete beta function and polygamma functions. Would it be of interest to incorporate some of these functions into openlibm? The whole library should be in the public domain.

[johnmyleswhite]

I'm certainly interested in the polygamma function.

[dmbates]

Have you checked the license, if any, under which this software is released? Fortran routines from 1990 described as being "made available, without cost, to the scientific community" can generally not be incorporated in an Open Source project covered by a license like the MIT license. Unfortunately numerical analysts of that period thought that phrases like "not for commercial use" or "public domain" were more meaningful than they are. A phrase like "not for commercial use" or "scientific community" is risky. It can taint the whole of Julia.

To be safe the software should have explicit copyright and explicit license conditions and I suspect that this software does not.

[andreasnoack]

I cannot find any license statements and I don't think there is copyright on it because it is produced by US Government officials and made available to the public.

Also, the author, Alfred H. Morris, Jr., states in the documentation that

"Since the beginning of the development of the library, no proprietary or otherwise restricted codes have been permitted in the library..."

"...Restrictions on the use of any library can severely impair its value, both for theoretical purposes (where the source codes are frequently of prime importance), and for general use in applications."

Personally, I feel pretty comfortable about the intentions, and that "scientific community" is not a restriction to non-commercial use, but it difficult to settle. Unless someone in the Julia community is a nephew of Leon Panetta I don't think the US military want to spend time on this copyright issue (I have emailed the US Navy). I have also emailed the University of Alberta and asked them.

[pao]

If it's a work of civil servants created for the US government, it cannot be copyrighted. You won't find a license, since without a copyright you cannot assert or enforce one. The US government restricts software distribution by other means, and if this is sitting on a public-facing Canadian webserver it's clear that this isn't ITAR, SBU, or classified. The release markings in the documentation "Approved for public release, distribution unlimited" are also a good hint.

[stevengj]

Unfortunately, government contractors (this includes a lot of the national labs) are not under the same restrictions regarding copyright. Without a clear free-software/open-source license, this is a minefield.

[pao]

I was very specific to say "civil servants" for a reason.

[andreasnoack]

I have been writing different people trying to find out the status of the software. The University of Alberta knows nothing more than what is on the homepage. Then I wrote NSWC Dahlgren trying to get a statement about the copyright status. After some mailing back and forth, it ended up with a non-answer and a reference to The Defense Technical Information Center. They again said that only NWSC can decide the status of the work.

It triggered some counterintelligence that I asked the NSWC as a foreign national so maybe the they will be able to consider a request from a US national writing from a US institution.

[StefanKarpinski]

Ah yes, Denmark, arch-nemesis of the United States.

[StefanKarpinski]

@pao, you willing to inquire? Might go better coming from an official-looking governmenty email address. Either that or I guess I could bust out my MIT address. Although with a name like mine they might not believe that I'm a US citizen either.

[pao]

I don't happen to have one of those working at the moment--my NASA account credentials have been wonky for a while (and it's not keeping me from getting work done, so fixing that is a low priority).

There might be some trees I can bark up, though.

[pao]

One could also check the report on DTIC and verify that A. H. Morris (or another actual civil servant) indeed programmed all the routines you're interested in. The introduction does imply that some of the routines come from other places, so the whole work is unlikely to be completely verifiable as clean.

[andreasnoack]

@StefanKarpinski ...against such enemies you have to protect the copyright status of the polygamma function. National security.

@pao I had a closer look at the manual and actually the author is stated for each subroutine. Morris did most of them. However, e.g. the polygamma function is based on code written by Donald E. Amos, Sandia Laboratories. It is a shame that Morris didn't think about formal licensing of these. The statement cited above indicates that he thought about the collection as free software.

I guess that only ACM would possibly complain about interpreting the NSWC library as free software because some of the subroutines are published in TOMS. Even thought the copyright is not transferred to ACM when the code is written by US government employees, it seems as if ACM "forget" to state that in the publications. Maybe if ACM is asked to specific algorithms they would answer if they have the copyright or not, but I think the best solution would be an answer from NSWC.

[ViralBShah]

I believe ACM does not get copyright for TOMS codes. There is a long thread somewhere about AMOS.

[andreasnoack]

I cannot find the thread, but at www.netlib.org/toms/ it is written that

"Use of ACM Algorithms is subject to the ACM Software Copyright and License Agreement".

But that is not the whole truth as e.g. Harris' work cannot have copyright.

[stevengj]

@alanedelman, do you think we can get MIT to make a formal request about the copyright status of this, in order to get the NSWC to pay attention?

[alanedelman]

Let's talk about who could even ask, and what is the desired response.

[ihnorton]

It might be possible to do this as a Freedom of Information Act request. This is legally required to generate some formal response within 20 days, although the response may be "we need more time to respond".

The tricky thing is that an FOIA request must reference specific records, not general questions. However, it might be possible to trigger a copyright determination as a byproduct of a request for the original code - there is some FOIA precedent for requesting documents even if they are otherwise publicly available. FOIA offices are required to determine copyright status before releasing documents. However, source code is not necessarily covered by FOIA, and that determination is at the disgression of the agency. One example of success with this route is the OpenVISTA project which has received source code from the VA electronic medical record system.

[cmcbride]

Any headway on this? My understanding is that this is holding up JuliaLang/julia#4301 and it'd be really nice to have some of those special functions in Base.

[jiahao]

Not really. Last I tried, MIT's legal office gave me a generic answer about released work done by public servants being automatically in the public domain, and promised to look into NSWC specifically, but never got back to me.

[StefanKarpinski]

I think that we should go ahead on the assumption that NSWC is fine to use.

[alanedelman]

sounds right

On Mon, May 19, 2014 at 10:16 AM, Stefan Karpinski <[email protected]

wrote:

I think that we should go ahead on the assumption that NSWC is fine to use.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/20#issuecomment-43509424
.

[johnmyleswhite]

Sounds like it's time to go ahead with this.

[ViralBShah]

This discussion is perhaps relevant in the openspecfun project, and not openlibm. Closing it here, since it has been discussed elsewhere too.