Re: slightly off-topic: support open source for publically-funded research

[Alex Lancaster]

[... I have only posted this follow-up to swarm-modelling, since it's
less off-topic than for swarm-support.  ;-) A quick disclaimer (which
I should appended before): my first post of the petition and my
subsequent comments are my personal opinions only and do not
necessarily reflect the opinions of the board of the Swarm Development
Group, of which I am a member, or of my employer, the University of
California, Berkeley...  There, I feel better now ;-)]

>>>>> "GR" == glen e p ropella <address@hidden> writes:

GR> Well, I'm sure this isn't the place to argue this; but, I'd just
GR> like to point out that agencies like the NIH, NSF, DARPA, etc.
GR> are funded by US taxpayers, not global contributors.

GR> Having the US publicly funded agencies force software created
GR> under their auspices to use open source licenses would
GR> subsequently force US citizens into donating their money toward
GR> helping other countries remain competitive with the US.  It's
GR> possible that this is unconstitutional.  Of course, there's an
GR> obvious analog to this in published research.  But, unless all
GR> research conducted with such funds is *required* to be publicly
GR> available, all software should not be so required.

You have a point, and purely as a legal issue, it would be hard to
argue that US taxes should be spent *directly* on software that would
be useful to other countries.  However, as a matter of empirical fact,
there is already a tradition of this in US research (as you rightly
point out there is an analogue with "published research", but there
are more direct examples of software, which I list below).  Also in a
purely economic sense it can make sense to make the code available to
a wider audience, since modifications made by non-US citizens are
required to be contributed back (at least if a strong `copyleft'
license like the GPL is used).  Examples:

(1) The Internet.  Although developed by the US-funded DARPA (then
    ARPA), virtually all the enabling technologies/standard were made
    immediately available to the world-wide engineering and scientific
    community (at least once the basic system was up and running as
    NSFNet).  Then, the Europeans at CERN (such as Tim Berness-Lee)
    developed HTML and HTTP, contributing the `killer app' of the net,
    "the web".  

    It can be argued that the resulting international collaboration
    resulted in building a robust network more quickly than if it had
    been the US alone, ultimately leading to a platform that resulted
    in economic benefit to the US.  Imagine a parallel universe where
    the protocols like TCP/IP and the basic DNS routing system were
    not made available across the world, we may have been in the
    paradoxical situation that the US would still have an old-style
    Internet (remember `wais' and `gopher'?), but no applications; and
    Europe would have the applications, but nothing to run it on.

(2) The Linux kernel.  Nonwithstanding Richard Stallman's, arguably
    more significant "GNU" component of "GNU/Linux", the Linux kernel
    was developed by Linus Torvalds, who as a Finnish citizen was not
    funded by US taxpayers, yet made code available that has been
    incorporated into a system that has resulted in direct economic
    benefit to the US.  Ultimately becoming the centerpiece of one of
    the US behemoths corporate computing strategy, that being IBM.
    
    This illustrates a kind of international quid-pro in open
    source/free software, some components may be developed inside the
    country or outside, but all benefit, especially those who invest
    the most (since they know the software better than so-called
    "competitors").  (Software can be considered a "nonrivalrous"
    resource in this sense, because my having access and use of the
    code does in no way diminish your access and use of the code.¹)

    Again, as parallel universe thought experiment: imagine what might
    have happened to GNU/Linux if the "GNU" component was forced to
    only be available to US citizens (leaving aside purely practical
    considerations such as enforcement).²

(3) GenBank/NCBI³.  Much of the human genome project research
    (including underlying software such as BLAST) was underwritten by
    US taxpayers, yet GenBank is available to all.

    Separate national projects that did not allow such information
    sharing would have most likely meant that the full sequence would
    not have been available for many years, would have necessitated
    more public spending and would have prevented the explosion of
    genome oriented industries and pharmaceutical benefits to the US
    that are and will begin to flow from this work.

(4) Funding foreign students.  An issue close to my heart ;-) Being a
    foreign student, I am funded indirectly via the NIH (and earlier
    in my incarnation as a Swarm developer at the SFI, I was being
    paid by a pool of funds that ultimately derived from the NSF and
    JWAC).  

    US agencies funds labs that have foreign students in them, again
    because of the quid pro quo situation: the investment pays off for
    the US because US research is forwarded, they have access to minds
    from the world over (for not a lot financial outlay, I might add
    ;-)), and the foreign student gains access to unparallelled
    educational resources.  Again, without trying to sound too much
    like a market-troid: it's a non-zero sum, `win-win' situation.

So even without invoking moral arguments like "it's the right thing to
do", or that "science is a global enterprise" (both of which I happen
to believe), it's possible to build a case that publically-funded
research made available to all (including non-US citizens) via open
source licenses, is ultimately completely in the US national interest.

One of my fears in this post-September 11 world the US will start to
retreat from it's existing commitment to free and unfettered global
scientific information sharing, and start to define "national
interest" a lot more narrowly than in the past.  It's an
understandable, if unfortunate situation.
 
GR> If the petition is re-written as "encouraged to be free software",
GR> then everything seems a bit more valid.

I agree that the wording could be tightened up here.  It's not 100%
clear to me that it should be an absolute requirement for all US
govt. funded research should be made available, the individual
researcher(s) and university should probably have some role in that
decision.

But, on balance, I think the goal of making it the "default" mode for
software developed by public money for scientific purposes (rather
than the existing "default" that it be closed or proprietary) is a
laudable one.

GR> Besides, the petition is specious because open source is not
GR> "...the software equivalent of peer-reviewed publication of
GR> research results".  This is mixing metaphors.  Open source is the
GR> equivalent of standardized tools in scientific research.  The
GR> software, itself, is not the research.  The software is one (and
GR> only one) description of one of the pieces of experimental
GR> equipment used.

>>>>> "MD" == Marcus G. Daniels <address@hidden> writes:

MD> In a peer-reviewed scientific publication, I don't believe
MD> agent-based models of any complexity will be described in
MD> sufficient detail to reliably reproduce results.  

[...]

MD> In these complex cases, the implementation of the model is
MD> not just an instrument but a crucial, bottom-line representation
MD> of the model.  Much the flexibility of agent-based modeling comes
MD> from the fact you can look at the dynamics of mechanisms that are
MD> otherwise hard to represent using conventional mathematical tools.
MD> This flexibility comes at the cost of clarity, but it may be worth
MD> it.

I agree with Marcus' point here.  It's not good enough (at least with
the state of the art in agent simulation) to produce the spec. and not
the code.  If everybody used the software equivalent of the "Tektronix
XYZ oscilloscope", then it might be, but we're not there yet.
(Incidentally the goal of being able to have a "higher-level"
specification is a laudable goal, the agent simulation equivalent of
Donald Knuth's "literate programming" idea, "cweb", comes to mind).

Even without the "code is the specification" equivalance argument, the
software reuse argument, is probably even stronger.  The ability for
other universities and researchers to re-use other tools could
ultimately save the taxpayers many more dollars.  Recall also, that
with open source software, if the university owns the copyright, it
can still relicense it under a commercial license, for those who wish
to make it part of proprietary product, and gain revenue for the
university and the individual researcher(s) that way.  The point being
that the original code will still be available in to the public who
ultimately paid for it.

Alex

¹ See Stanford law professor Lawrence Lessig's recent book, "The
Future of Ideas", for a more fully articulated vision of the
importance of global information commons to innovation.

² Actually, much to Stallman's delight, we might all be running
GNU/Hurd, Hurd is a kernel developed by Carnegie-Mellon University,
but that's another story entirely...

³ National Center for Biotechnology Information:
http://www.ncbi.nlm.nih.gov/

-- 
   Alex Lancaster * <address@hidden> * www.santafe.edu/~alex 
Dept. of Integrative Biology, UC Berkeley (ib.berkeley.edu) * +1 510 642-1233
    & Swarm Development Group, Santa Fe, New Mexico (www.swarm.org) 


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================