And one more (Re: yet another talk submitted)
Hi,
On Thu, Apr 29, 2010 at 02:24:23PM -0400, Yaroslav Halchenko wrote:
> On Wed, 28 Apr 2010, Kumar Appaiah wrote:
> > In addition, if folks here have a request for a talk on a particular
> > topic, which they would not be able to present themselves, it would
> > still be nice to convey the suggestions to the list and/or Michael,
> > so that others who might like the idea could take it up, and
> > potentially talk about that subject.
>
> Although not requested, I've decided to share the abstract I've just
> submitted. I wanted to get your feedback (does it sound worth
> "lecturing" or may be some kind of "discussion" would be more
> appropriate).
Here is my own submission -- related to Yarik's in that it is a
concrete example of (successfully) promoting Debian in a particular
field of science -- neuroimaging research.
------------------------------------------------
Debian: The ultimate platform for neuroimaging research
Over the past decade the neuroimaging research community has
fortunately converged on the open source software development model,
with the vast majority of all widely used applications and libraries
being available as source code, and a substantial proportion covered
by free software licenses. Today, there exists a large code base for
data collection (e.g. behavioral psychological experiments), data
analysis (e.g. of functional magnetic resonance imaging; fMRI), and
data visualization that is productively used in everyday research
activities. A representative list of available tools is provided on
the NITRC portal [0].
Initially, most of these tools were mere by-products of actual
neuroscience research projects, developed by students, and scholars
who are, in general, not trained software engineers. As a result,
software development in this field still differs significantly from
established good practices in the free and open source software
(FOSS) community. Many projects start without a clear deployment or
management concept, and due to lack of man-power are later on forced
into an \textit{ivory tower development} model -- restricting the
``supported'' environment as an attempt to reduce the required
maintenance effort. They try to decouple themselves from ongoing
developments and include specific versions of all external
dependencies into their source distributions. The result is too often
a fork that is no longer updated with bugfixes or enhancements. Many
useful projects die because it becomes too expensive to update them.
However, it is hard to blame the respective developers, because the
sheer number of existing combinations of operating systems, hardware,
and library versions makes it almost impossible to verify that a
particular software is working as intended. Restricting the
``supported'' runtime environment is one approach of making
verification efforts feasible. On the other hand, development in the
free software community sometimes proceeds at an enormous pace,
and many projects do frequent releases with sometimes
backward-incompatible changes. Forcing a dependency on an outdated
release shifts the burden of maintenance and deployment onto users, as
it gets increasingly difficult to have older releases available. On
the other hand, continuously updating software to the latest
developments is a time-consuming task for which there is no immediate
scientific benefit.
I will argue that Debian is the ideal solution to this problem. This
talk will provide an overview of current projects and individual
efforts within Debian that help researchers to maintain a
fully-functional Debian-based environment for neuroscience research
with minimal effort (e.g. Debian Med, Debian Science). Moreover, it
introduces NeuroDebian [1] -- a platform specifically targeting the
neuroscience community. I'm going to show several case examples of
packaging efforts that illustrate typical problems, such as
non-commercial/non-standard licenses, non-existing upstream bugtracker
and unavailable version control systems. Furthermore, I will offer
some evidence that a substantial migration towards Debian is already
under way. Popularity statistics and personal communication indicates
that a growing number of researchers see it as 1) making neuroscience
software (e.g. for medical imaging) accessible to a larger user-base,
and reducing the required maintenance effort on the user side, 2)
improving software quality, and 3) enabling developers to spend more
time on developing new scientifically relevant features. The talk will
conclude with thoughts on what Debian could do to further facilitate
its adoption as the ultimate platform for scientific research, such
as tailored guidelines for deployment and project management,
possibility to run complete regression test suites, as well as
reliable usage statistics to justify funding project funding.
[0] http://www.nitrc.org
[1] http://neuro.debian.net
------------------------------------------------
--
GPG key: 1024D/3144BE0F Michael Hanke
http://mih.voxindeserto.de
Reply to: