CDDL Project Leader on the CDDL

SunFan writes "Claire Giordano, Sun's lead on CDDL development, gives the inside story on how the new license was developed. She discusses how people within Sun debated the various licenses, including the GPL, and shows why the GPL, BSD, and MPL licenses were found to come up short in meeting the needs of Sun's broad customer and developer base."

Comments

MPL-style license? try LGPL.

[ebyrob]

What the author of "Failed as in succeeded wildly." meant was LGPL-style licenses. LGPL has been around a lot longer than MPL and actually fits what he describes.

Real Problem

[Anonymous Coward]

It's not GPL-compatible. I appreciate you want to be able to mix in proprietary code. If you need that, you can do one of two things. You can use LGPL, or you can include a clause "May also be sublicensed under the GPL, version 2.0 or newer." The only thing you lose is the patent litigation protection. You can also grant yourselves rights to change the main CDCL without alienating the mainstream FLOSS community.

Either way, a large number of people, for whatever reasons, distrust Sun. Releasing under a license that is widely percieved to be incompatible with the greater free software community is fanning those flames. Most people percieve this as a simply ploy to make it impossible to move code back and forth between the Linux kernel and Solaris. That's not really big on freedom.

One of the great parts of free software is rapid development. About half of my apps are built by glueing together code from other people's programs. Making Solaris not compatible with the GPL breaks that, and it's not good for building a developer community around Solaris.

I wish you guys the best of luck, but I think that given a GPL-incompatible license, you'll have a really hard time building a large developer community.

With a GPL-compatible license, if Solaris got to be better than Linux, virtually no one would think twice about switching from Debian GNU/Linux to Debian GNU/Solaris.

Also, notice the trend. Virtually everyone (Netscape, TrollTech, etc.) starts with their own crappy license, and eventually, switches to a dual-licensing model that includes GPL. Follow the lesson, and start off on the GPL side from the start.

Re:Real Problem

[hacker]

"Non-GPL code becomes GPL. It's like an irreversible chemical reaction, unless you happen to own the copyright to all the code to begin with."

I hear this "viral" kool-aid all the time regarding the GPL, and I do find it funny every time it comes up. The GPL is not viral. Repeat that to yourself. Call the FSF and ask them.. No really, go ahead, I'll hold.

I know this intimately, because we're fighting a GPL case now (4+ years and counting) against a company that took our code, removed our names, called it their own and ship it without the license, including linking parts of it to their own products. We're FSF-backed (they've appointed us an incredible attorney), and we've gone over every angle of this hundreds of times.

If the "if it touches the GPL, it must be GPL" FUD was real, do you truly think companies like Oracle, VMware, and thousands of others would deploy their binary-only products on Linux, which links to... guess what, tools, libraries, userspace programs covered under either the GPL or the LGPL.

It goes like this. Watch carefully now:

1. You write a proprietary product
2. You want to add a feature, so you take some GPL code and add it to your proprietary product
3. Author of the GPL code finds that you are using his code, and requests one of two things:
   1. The full source code to the pieces that link or modified his code, or...
   2. That you no longer have the rights to use his code in your proprietary product
4. Your rights to continue to use the code under the GPL are immediately revoked.

At this point, you are no longer covered by the GPL license, and any further use of the GPL code is now a United States Copyright Violation (and possibly a Lanham Act violation, which is defined as "...false designation of origin").

So to get yourself out of this mess, you have a few options, all of which require that you stop using the GPL code in your product (well, except one; see below). There is no way to continue to use GPL code in a violating way, period. Many people believe you just pay a fine and business goes on, but it doesn't work like that. You're still guilty of the violation.

1. Settle with the developer for an undisclosed fee, a fine.
2. Go to court, and let the judge decide what your penalties are (US Copyright violations carry anywhere from $20k to $200k in fines, per-incident). They'll most-likely issue a blanket settlement in the sub-million dollar range, because its hard to determine "damages" of GPL code when its given away freely, and downloaded/shared all over the place.
3. Encourage the developer to relicense the code to you under a non-GPL license

Again, at no point does YOUR PRODUCT somehow become GPL code. It just doesn't happen, ever. There isn't one single precendent of this happening, because that's not how GPL violations are handled.

So please, in the future, do some research, read the GPL (actually READ it), and call the FSF or the EFF if you wish, and find out the real facts. Stop spreading FUD, you're only making the problem worse for yourself.

Lame Excuse

[sbszine]

From TFA: On Day 1 of OpenSolaris, because some OpenSolaris IP is encumbered by other companies (example - 3rd party drivers), we're going to have some source files in the kernel that will remain proprietary. Hence GPL was out of the running.

The whole point of opening the source and creating a community is so that people can develop the things they need, free of the problems proprietary code brings. I find it hard to believe that much Solaris kernel code belongs to anyone other than Sun.

Linux started out with very few drivers, but now supports most common hardware. Sometimes this support takes the form of binary or wrapped drivers, but that hasn't prevented Linux from remaining under the GPL.

This is just more of the usual Sun guff.

Re:And what lame excuses they used

[Anonymous Coward]

Why not do the same thing as the Linux community, and release it without those drivers until someone writes an open-source replacement.

Because, ultimately, no one gives a rat's ass about a proprietary driver. If it works, that's great. If it's open source, well, that's nice, but working is better.

Re:Wow were SUN

[Martin Marvinski]

OpenSolaris is completely open! You can create your own OpenSolaris Distro! Casper H. S. Dik, one of the OpenSolaris community advisory board members talks more about the CDDL. He basically states the CDDL is the MPL without some of the restrictions..

An example he gives is the right for the Mozilla foundation to revoke the MPL and the requirement to have lawsuits settled in California (which would be bad for international users). That is why SUN didn't use the MPL.

SUN enginner Alan Coopersmith points out in comp.unix.solaris that anyone can create an OpenSolaris distro.

http://groups-beta.google.com/group/comp.unix.so la ris/search?group=comp.unix.solaris&q=Alan+Coopersm ith+cheapbytes+Opensolaris&qt_g=1&searchnow=Search +this+group

Re:Wow were SUN

[Martin Marvinski]

One more thing. Joerg Schilling, famous for creating Linux's cdrecord program is creating his own OpenSolaris Distro called SchilliX. http://schillix.berlios.de/ Blastwave.org, a website that contains free software for Solaris users is also creating an OpenSolaris distro according to the regulars of the usenet group comp.unix.solaris. A benefit of OpenSolaris is that it will have a proper cdrecord functionality. Again SUN Engineer and OpenSolaris community advisory board member Casper H. S. Dik says this about OpenSolaris compared to linux. In message ID Casper Dik writes: "I think [Joerg Schilling (author of cdrecord)] prefers neither but rather has a proper USCSI interface; note that this isn't really a "IDE SCSI" emulation layer; ATAPI is SCSI-over-ATA. " The problem with the linux kernel is that a scan by cdrecord may not find all of your cdrecord devices because of the changes in the 2.6 kernel. Even Alan Cox mentioned that it was not optimal in one of the long linux.kernel threads. :-)