John H. Robinson, IV wrote:
Not true. OpenSSL is *very* hurt by this as their code cannot be used in
any GPL application, without a special licensing. GNUtls does not suffer
from this, but it is not as featureful as OpenSSL yet.
Oops. I meant *openssh*, mea culpa.
What license is OpenSSL under? I thought it was Apache license? I
don't recall the Apache license being GPL incompatible, but I am willing
to be corrected.
Howso? I can see problems where you want to *change* the license, or if
you want to appropriate the code in your own, prorietary (read: non open
and free) project.
Short of those two situations, what problems have you experienced?
Computational geometry subsystems are a good example. There are quite a
few computational geometry subsystems available--all under various
licenses which restrict redistrbution.
I have a new algorithm for doing VLSI compaction, layout, etc. but it
uses needs computational geometry subsystem. However, most
computational geometry subsystems tend to be a couple years out of date
as there isn't tremendous call for them. So, I have to take the code
and recompile it myself on modern machines.
Now, I need to hand my code to a subcontractor to check the chips too.
So, what do I do about the computational geometry engine?
Since, I can't link with the computational geometry engine, I can't hand
the subcontractor a binary. I have to break the code apart, but I can't
run them separately as there is too much data to move. So, I have to
create a full installer that handles both my code and the computational
geometry code separately. Normally, by this time, I get pissed off and
write my own crappy computational geometry code so that I can actually
link with it. Sure, it's not as good as the released library, but it's
good enough.
So, the released library doesn't improve because I have no incentive to
do so. My library is sufficiently underfeatured that it is not worth
releasing. The world at large loses out.
GPL is nice when there is an active critical mass, but not a huge
number. BSD is nice for the very small niche stuff which tends to
languish years between uses and the very large stuff where getting
everyone to use it is more important than preventing change(the TCP/IP
stack, for example).
That's my experience, YMMV.
-a
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
