[BioPython] Phylogenetics trees

Andrew Dalke dalke@acm.org
Thu, 31 Aug 2000 12:34:49 -0600


>However I would really rather keep all the code in mavric under the GPL. 
>At the risk of raising hackles I must admit I like the requirement that
>modifications to the code be made available-- it appeals to my ideals
>about how science should be done, i.e. free and open access to ideas and
>improvements (sorry to be so melodramatic :) 

I can understand the desire that modifications to the code be made
available.  I've also released code with that requirement.  However,
I did it under the LGPL, which doesn't have the clause mandating that
all other code be under the same license.  At the very least, this
allows use with other free, but incompatible, licenses.

It's also not true that changes to the code will be distributed even
if the program is available for public use.  I've given the example
of web sites which offer access to modified programs with special
features.  The code is GPL'ed, but the use doesn't fall under the
licence's definition of distribution.  Also, if someone really wants
to turn a GPL library into a LGPL-like use (where their code is not
affected by the GPL restrictions) then it's pretty easy to put a
CORBA or SOAP wrapper around the library and use it from elsewhere.

>I could be conviced to re-license the phylogeny code and fold it into
>biopython if anyone can tell me how this would benefit the project in ways
>it could not as a separate 'third-party' module?

I don't think there is a large benefit.  Of course, one-stop shopping
is always nice, but then there are people who would say the distribution
is too big because it includes things they don't need.

>Also, if the the phylogeny code did become part of biopython, does this
>mean that mavric would thereafter depend on users having biopython (for
>the phylogeny libraries)?  Sorry to be dense here. 

Well, why wouldn't they?  In fact, I'm going to petition Guido and
insist that biopython be included in the core Python distribution.
Or not :)

No, you are correct.  I don't see there being an advantage now for
the libraries to be integrated into biopython, given that there's only
a single package using them.  The question will be, at what point will
people want to get access to those libraries for their own code?  How
tightly bound is it to what you are doing?  (Sorry, I haven't looked
at your code, so I'm only speaking generalities here.)

It's like Gtk, which was part of the Gimp, but spun off into its own
project.  When things get to that point, you'll probably maintain two
components (one for the tree manipulation code and one for the GUI
application), or someone else will make that library.  Only at that
point will there be the question of where the library will be maintained.
Odds are it will be like Python, where there are SIGs related to certain
projects, with some projects built off of others.

                    Andrew
                    dalke@acm.org