[Python-Dev] BSDDB3

[Steve Holden]

Gregory P. Smith wrote:
> 
> On 3/4/08, *Jesus Cea* <jcea at argo.es <mailto:jcea at argo.es>> wrote:
> 
>     That said, it is my aim to keep bsddb in stdlib, providing a stable and
>     featureful module. I think keeping bsddb development inside python svn
>     is not appropiate. Currently (I could change idea), my approach will be
>     keeping pybssdb as a separate project and sync with python SVN from time
>     to time. Mainly to take advantage of buildbot architecture and, of
>     course, to be able to release python with current bindings.
> 
>     Since I have no python commit access, this seems a sensible approach.
>     And I could do frequent pybssdb releases (let say, every couple of
>     months) without waiting for a full python release (current approach).
> 
> 
> That makes sense.  I also agree with Neal's comments, merging back into 
> python in reasonable sized chunks is good.  Don't worry about commit 
> access for now, I'll do the initial pybsddb back into python trunk 
> merges until we get that setup.  I've merged the python trunk changes 
> that others have made back into the pybsddb tree.
> 
>     PS: I have tried to sign the Python Contributor Agreement, but I am not
>     sure about current pybsddb license terms. Help welcomed.
> 
> 
> The current bsddb license first and foremost is the Python license.  If 
> I read the comments in the _bsddb file correctly I believe you could 
> also call it a MIT style license.  Keep things simple, just write 
> "Python License" on your contributor form and submit it.
> 
Unfortunately that won't do it. At present the PSF can only accept 
contributions under two initial licenses: either the Academic Free 
License v 2.1 or the Apache License v 2.0. See

   http://www.python.org/psf/contrib/

This is because only these licenses have been approved by attorneys as 
giving the PSF the necessary unencumbered permission to relicense for 
distribution *under* the Python license.

So Jesus was right to be concerned about licensing. I know it's a pain, 
but there are reasons. I don't see anything in the file to stop _bsddb.c 
being licensed to the PSF under either approved initial license. The 
licensing FAQ

   http://wiki.python.org/moin/PythonSoftwareFoundationLicenseFaq

explains why you can't just contribute code "under the PSF license", and 
the contributor agreement requires that each contribution should include 
the contributor's valid copyright notice and the phrase "Licensed to PSF 
under a Contributor Agreement." This condition appears to be more 
honored in the breach than the observance though, since the phrase 
currently only seems to appear five times in the whole source unless I 
need to practice my grep-fu.

As to the library that _bsddb wraps, I have not checked its licensing 
conditions and so cannot say how it is licensed to the PSF for 
redistribution, nor do I know whether Oracle's acquisition of SleepyCat 
will affect future versions. Bet it gave the MySQL guys some sleepless 
nights, though.

regards
  Steve
-- 
Steve Holden        +1 571 484 6266   +1 800 494 3119
Holden Web LLC              http://www.holdenweb.com/