httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ralf S. Engelschall" <>
Subject ASF and licensing (was: Re: GNU Portable Threads)
Date Tue, 06 Jul 1999 07:49:56 GMT
[Hmm... I've once again written more than I wanted to write. And I'm now very
afraid that some of us will again feel offended by my statements because I
think too differently here than the group as a whole. Nevertheless I want to
share my opinion, because I think the topic is important. I've read the above
text three times and think my used English is not really offending, but I've
to admit that I'm not sure. At least I don't want to directly offend some
particular members of us, I just want to point out that there can be a
different licensing point of view which hasn't be too unreasonable.  So,
please be really smart in reading the above text and do not weight all of my
writing word by word. Try to recognize my general expressed opinion and do not
quote my exact wording.  Thanks. Sometimes I wish I had a native speaking
secretary... ;)]

In article <> you wrote:

> Let's talk about this sanely for a second.
> [...]

First, thanks Brian that you try to summarize the situation and try to find a
way for us. You know, for independent pure private hackers like me there are
reasons to use LGPL and not BSD-style licenses (we want that our work and
derived work is _kept_ free).  On the other hand it's clear that the guys from
the commercial camp have to dislike it to some degree (because LGPL or even
GPL applies restrictions to them they usually would like to not have).
Nevertheless it's important IMHO for the success of the ASF that we find a
reasonable approach to make both camps happy and allow them to co-exists under
the same ASF umbrella.

> [...]
> By my reading of the LGPL (and Bill, correct me if you think IBM has a
> different opinion), we can have the end-user link Apache against LGPL'd
> libraries.  We can also (separately) distribute an LGPL'd package of all
> the essential LGPL'd libraries Apache would require during compilation,
> even with whatever mods we needed to make for whatever reason, all again
> under the LGPL.

That's how I also read and interpret LGPL myself. And from the known arguments
of RMS we can conclude that he interprets it this way, too. But as I will
mention, what RMS says is not really important for us.

> However, we *have* to treat (Apache & other non-LGPL code) and (LGPL'd
> code) as separate "works" requiring separate distribution.  Where I think
> IBM and other large institutions have a problem is with trusting Stallman
> not to sue them over a differing idea of what a "work" is.  

How RMS interprets it personally is not directly important for us unless we
talk about a program/library which is under GPL/LGPL _and_ at the same time
copyrighted by the FSF (only a few things are of this type). When you've a
GPL/LGPL'ed package (including my GNU pth), IMHO and AFAIK the one you have to
care about is _only_ the copyright holder who placed the package under

> [...]
> So, there's the issues as I know them.  If others have any, please bring
> them up.  I think there *is* a way to reconcile the use of LGPL'd
> libraries with the commercial interests lined up behind Apache, although
> it is somewhat inconvenient.  

Yes. But please let us keep in mind that it is _intentionally_ inconvient _per
default_ IMHO. I'll try to explain it from my point of view:

In the commercial world when you buy a product from another company you
get a _very_ restrictive license which usually only allows you to _use_ the
program. When you really want to _incorporate_ the program in your own work,
you've to get a special license from the vendor (which usually costs a lot
more $$). Then you usually get a few object files and can link your product
with it. And when you really want the source you either cannot get it at all
or have to pay even more $$. Everyone accepts this practice in the commercial
world. That's just how money is made there.

It gets really interesting when the commercial world wants to use free
software in their products. They already can use it for free. And as long as
they follow a few restrictions (which mainly make sure the software cannot be
relicensed and this way evolve into non-free status over time) they can even
include the stuff (even as source code!) in their own product. But they
_still_ complain, even they already have a _better_ situation than with any
commercial stuff they're usually used to. Isn't this interesting?

What they moreover totally forget is that there is no difference for them
compared to the commercial situation: when they _insist_ on a better licensing
situation for them, all they have to is the same they would do for a
commercial product they want to include: They contact the vendor (in this case
the author and copyright holder) and arrange either a few license exceptions
for them or even a completely different license for their individual copy of
the package.  THAT'S ALL! And that's what was done, is done, and will be
done for lots of GPL/LGPL'ed packages, btw.

Free software which is placed under a distribution license has to protect
itself perfectly for the _default_ case with the help of the license. So it's
clear that some licenses like GPL/LGPL are very restrictice.  They have to be
this way in order to achieve their goals (mainly to keep the software free).
When one needs special conditions, it is totally acceptable and common
practice to arrange this with the author.  The same way one does this in the
commercial world for other products. The interesting aspect is just that
people seem to forget this fact.

So, to give an example: When IBM wants to create a commercial closed product
out of Apache-pthread plus GNU pth, they can do the following: When it's
acceptable to them they can just silently fulfill the LGPL clauses of pth and
everyone is happy.  When IBM thinks this doesn't fit very well into their
situation (perhaps it's still to inconvinient, violates some of their
marketing strategies, or whatever), all they have to do is to contact the
author and arrange a special license for their situation.  When it's just a
matter of allowing them to distribute the stuff as a single tarball instead of
separate tarballs, no author will usually complain. IBM usually would
immediately and without charge get such a LGPL exception clause for their
product.  When they want more (perhaps that they want to keep their
modifications secret) they need a special individual license. But they can get
this, too.  Ok, perhaps then no longer without charge, of course.  But hey,
that's no way different than what is common practice for commercial software.
So why do they expect it differently all the time?

So, about what I'm all the time very surprised is the fact that the commercial
world state they _support_ and do not _exploit_ the free software world but at
the same time allow theirself to complain about free software licenses when
they do not per default allow them do whatever they want with the free
software. They seem to assume that just because it's open source and doesn't
cost anything to be used, they also can create a commercial product out of it
without having to carefully make theirself sure the stuff is kept free
software. Authors like me usually don't try to prevent the creation of
commercial products out of our own work. No, it's ok when someone else makes
money out of our work.  But what we usually at least want is that our stuff is
kept free. That's a great difference. So, it's really interesting that with
free software the commercial world often implies that the license deals, they
are usually used to, do not exist.

Sure, whether such approaches are reasonable for the ASF is a different kind
of question. Perhaps we conclude as a group that one of the ASF major goals is
to explicitly support the creation of closed commercial software packages
based on the ASF's work (notice: not the commercial aspect is important here,
it's the aspect of the closed/proprietary nature of the product).  When one
assumes this it might be not necessary to first restrict the commercial world
and then give them better priviledges. Then it can be acceptable that we _per
default_ (for instance by using a BSD-style license) allow them to do whatever
they want with our stuff. 

But I have to admit that such an approach doesn't make me personally really
happy.  Because I personally spend _LOTS OF_ free time on coding without
getting any money, so I at least want that my work is kept free.  And when it
isn't I at least expect a similar compensation I would get when I had coded
the stuff in the commercial world. That's just fair. And when the ASF wants to
be fair it should support people like me. Because I'm sure I'm not the only
one who wants to work for the ASF but at the same time don't wants to be
exploited by the commercial world...

BTW, one more note because I know we will read it as one of the replies: I
know that most of you work in the commercial world and explicitly say you
_support_ the free software community by your _contribution_. That's great!  I
even doubt all of you that you would still contribute and this way support the
free software community even when you wouldn't work for and get paid by the
commercial world for your doing. 

Nevertheless the usual argument itself ("we support free software, so we can
expect a good license on it") can be considered a little bit unfair: you get
paid for your contribution mainly because the commercial company who pays you
benefits indirectly from the contribution. So the contribution-vs-own-benefit
itself is an acceptable balance here. But IMHO the balance is broken when one
at the same time requests a license which lets one to do whatever one wants
with the source in the commercial world.  Then the benefit is more at the side
of the commercial world and one could consider it some sort of exploiting

Currently when I understand us correctly the ASF as a group thinks about the
whole as following: We create open source software with a very less
restrictive license. This encaurages the commercial world to support Apache by
contributing and because they base their products on our work we indirectly
get even more of the market share. Drawbacks like the fact that there maybe
modifications and bugfixes in the commercial world we do not get back have to
be accepted with this approach. So as a whole one cannot say that this isn't a
reasonable approach. 

Especially most of us are happy with this, because they are already paid most
of the time exactly by those commercial companies who benefit on Apache and so
they are paid for what they usually would do in their free time. So it's clear
that we cannot expect these people to deny the requests of the commercial
world, of course. And the users? They are happy, although they know that it
could theoretically happen that Apache in the future is no longer supported by
the ASF and instead evolved into a commercial product only. And they also have
to accept that commercial variants of Apache exists which are "better" then
the official Apache.

But one could also do it differently: The ASF creates open source software
with very restrictive licenses like GPL/LGPL or even a similar own license.
This encaurages not immediately the commercial world to support Apache by
contributing, because they not immediately benefit from it (because they
cannot do whatever they want with the code inside their commercial product).
Nevertheless those commercial companies who contribute show _real_ support for
Apache. And those companies also have no problem in fulfilling the restrictive
license clauses to make sure the software is kept free although one makes
money out of it.

The remaining commercial companies who want to base their product on Apache
but don't want to follow the clauses of the default distribution license
contact the ASF and _pay_ the ASF for a special individual license situation.
In other words: the ASF licenses Apache to the commercial world for a good
price when they want a special licensed Apache. Not such a lot of money other
commercial software would cost, but at least money the ASF this way receives
and which is important for making conferences, etc. And the users? They are
_really_ happy, because they can be sure that even in ten years there will be
a freely available Apache. And even more: This way we both technically have
less problems (we can include things into Apache we else couldn't easily
include) and can host both the commercial and private hackers under the same
ASF umbrella.

Yeah, I know I'm very heretical with my opinion. But I'm really such
unrealistically just because I think the ASF could do it also differently?
And isn't my suggested approach the way both camps can be happy at the same
time? Hmmm... I would really appreciate when the world would be more easily ;)

                                       Ralf S. Engelschall

View raw message