[Zlib-devel] zlib and LZMA

[William A. Rowe, Jr.]

Hi - sorry these observations are a little late - must have had connectivity
issues the day I drafted it...

Mark Adler wrote:
> 
> I envision making zlib available with and without LZMA, so the  
> current deflate-only zlib under the current zlib license would still  
> be available.

I just spent at least 4 hours this week in part of a license audit; I'm
going to guarantee you less adoption and more moving away from this code
once you deploy a more confusing licensing scenario.

I totally agree that two packages won't harm your current users, but by
mixing up the licensing terms, I can warn you that enough of us out here
have lawyers who will just say 'this is unnecessarily complex, choose
another solution'.

For example, unless I heard otherwise from the ASF legal team, only the
deflate-only flavor of zlib would likely to continue to be bundled
in Apache HTTP server binaries that I create.  That would be a shame.

> However I would work with Igor (who is on this mailing  
> list) to make the license with LZMA as palatable as possible for all  
> users including commercial users.  The only difference will be that  
> the LZMA cannot be modified by the commercial user for use in their  
> product without providing those modifications back to us for open  
> distribution.

Please please please do this only using a conventional open source license,
since - again it's our laywers - are already versed in the particularities
and can give thumbs up and thumbs down, without huge investments in review
of YAOOL (yet another one off license).  Pick the existing one closest to
the desired goal.

It is a shame Igor hasn't recognized the success of the zlib development
model - more companies will invest more dollars and provide him -more-
feedback, not less, on optimizing his LZMA implementation under a license
totally usable by every community.  For patent protection, which is really
a much bigger issue than the whole copyleft debate, perhaps look at the
Apache Software License 2.0.  But of course, anything he does to ensure
that nobody plays patent games around his sphere of work will undermine
compatibility with the GPL which permits no other restrictions.  Question
comes down to which is the broken license, in this case, but it's an issue
that needs to be recognized and a flamewar not worth repeating.

Oh, if any one-off copyleft license is anything except GPL or LGPL, it's quite
easy to be incompatible with the GPL; another reason to use only an existing
license with both OSS opinion and FSF opinion already rendered.