[Insight-developers] Telcon notes, 6 July 2001

[Paul Hughett]

Notes from Insight Telcon, 6 July 2001:

1. Checking in data for examples and test program is becoming a
problem; CVS is not good at handling large binary files.  FTP is a
possibility but has potential security problems (passwords are
transmitted in the clear).  Using SSH 2 (and the sftp program in
particular) may be a good choice; Unix clients are definitely
available, and Windows clients are probably available.

2. ITK does not seem to compile correctly under the Solaris 5.0 C++
Compiler (6.0 is the current version).  We will try to get it to work
but may decide to not support 5.0.

3. The "make" and "make clean" targets sometimes fail because of old
configuration information left behind by previous compilations.  The
easiest way to really clean out a out-of-source build directory seems
to be to run "rm -rf *" in the build directory.  (Be VERY SURE that
you're in the right directory.)  There does not seem to be any
equivalent trick for in-source builds.

4. The multi-resolution mutual information registration code is
currently failing to compile on Linux because it uses more levels of
templates than g++ provides by default.  The build options will be
modified to fix this.

5. The RawIO package needs quite a bit of refactoring and rewriting.
We will provide a "just good enough" version for the upcoming release
and make the heavy modifications afterwards.

6. The names of the various releases have been clarified: The "private
alpha" is what we released earlier this year to a few close friends
and associates.  The "public alpha" is the release we are preparing
now, for distribution by invitation only.  The "beta" release is the
one planned for November 2001, which we will give away to anyone who
wants it.  The alpha releases are unsupported and largely
undocumented--whatever you see is all you get.  The beta releases
should include useful documentation and some minimal commitment to fix
bugs soon after they are reported to us.

7. Terry has emailed the proposed agenda for Seattle; comments and
revisions should be addressed to him.  We want to spent a lot of time
brainstorming and developing a solid plan for finishing up this
project.

8. Lydia has committed examples for multi-resolution mutual
information registration and for fuzzy connectedness and encourages
people to try them out.

9. Terry asked where were all the segmentation algorithms that
comprised such a large role in the original segmentation/registration
toolkit proposal.  He was answered that a lot of segmentation code
does exist, but isn't really called out in the documentation.  One
issue has been choosing the right packaging and granularity for
various simple but widely used filters.  Nevertheless, we do need to
compare what we've actually implemented to what was proposed, and fill
in any gaps.  Ditto for making sure we have all the common, widely
used operations in addition to the fancy, fun-to-implement algorithms.
This will be a topic for Seattle.


Paul Hughett