[Insight-developers] Ramblin' on: Another concern regarding IJ and ITK code contributions

[Torsten Rohlfing]

Greetings everyone --

I would like to add to the general discussion another issue that I see 
in the current model of ITK code contributions through the IJ. But to 
not create any unwarranted suspense, let me say one thing beforehand: I 
have, again, no idea what a possible solution might look like, and maybe 
I am the only one who even thinks that this is a problem in the first place.

Anyway, here we go: in my opinion, channeling code contributions to ITK 
through the IJ makes it very hard to create a cross-institutional 
collaborative effort for larger ITK extensions and for the 
implementation of new concepts.

Here's an example from my personal experience: the code and data 
structures for diffusion tensor images came out of a community 
discussion and at least three sample implementations (by Jeffrey Duda at 
Penn, Martin Styner at UNC, and myself). Before this effort, I didn't 
know Jeff, and I had no idea Martin was interested in DTI. So in the IJ 
submission model, each of us might have submitted their own 
implementation, with their own data structures, and some application 
code around them. Assuming that all three implementations would have 
attracted sufficiently many favorable reviews, it is unclear to me, how 
from there one would have gotten to a single "official" implementation 
(and when that might've happened).

A more recent example: I noticed two interesting contributions in the IJ 
-- a "general" image fusion framework ("GIFT"), and a contribution to 
create overlays of label maps onto corresponding intensity images. It 
occurred to me that it would probably make sense to unite these two in a 
common framework for image fusion (or, to quote a Monty Pythons sketch: 
"for putting things on top of other things").

It also occurred to me that there are probably quite a few people out 
there who'd be interested in such a general image fusion framework, only 
they're scattered across many different institutions and would probably 
not find each other easily. So in the best case, each of them will write 
an IJ submission with somewhat incompatible and somewhat redundant code. 
In the worst case, many of these folks will find the "entry barrier" of 
getting everything coded and documented and polished and written up for 
an IJ submission to high, and only a fraction of the possible 
contributions will actually happen. And then these would probably still 
be incompatible and redundant.

So that's basically my criticism and my question - it seems that the 
previously practiced model of ITK development would have accommodated 
this situation quite nicely, and in my own experience it did so for the 
DTI code. I am wondering if there are any ideas how one could maintain 
the undoubted benefits of the IJ submission model (thorough review; 
stable interfaces) could be extended to encourage, or at least enable, 
dynamic forming of cross-institutional development efforts that are 
beyond the capabilities of any one group alone?

Cheers!
  Torsten

-- 
Torsten Rohlfing, PhD          SRI International, Neuroscience Program
 Research Scientist             333 Ravenswood Ave, Menlo Park, CA 94025
  Phone: ++1 (650) 859-3379      Fax: ++1 (650) 859-2743
   torsten at synapse.sri.com        http://www.stanford.edu/~rohlfing/

     "Though this be madness, yet there is a method in't"

-------------- next part --------------
A non-text attachment was scrubbed...
Name: torsten.vcf
Type: text/x-vcard
Size: 366 bytes
Desc: not available
Url : http://www.itk.org/mailman/private/insight-developers/attachments/20060920/6ebb413b/torsten.vcf