Slashdot Mirror
Debian Kicks Jörg Schilling
An anonymous reader writes "Debian's cdrecord maintainers announced that they have had enough of Jörg Schilling and kicked his program suite cdrtools out of Debian, introducing a free fork of his no longer free cdrtools." I've put the message below, along with some other links.
So, why the fork? CD/DVD burning is a complicated business that needs a lot of knowledge, so forking such a big collection isn't a step to be taken lightly. It requires a lot of development effort that could be put to better use elsewhere.
In the past, we, the Debian maintainers of cdrtools, had a good and mutually cooperative relationship with Jörg Schilling. He even commented on Debian bug reports, which is one of the best things an upstream maintainer can do. Naturally, there were occasionally disagreements, but this is normal.
Unfortunately Sun then developed the CDDL and Jörg Schilling released parts of recent versions of cdrtools under this license. The CDDL is incompatible with the GPL. The FSF itself says that this is the case as do people who helped draft the CDDL. One current and one former Sun employee visited the annual Debian conference in Mexico in 2006. Danese Cooper clearly stated there that the CDDL was intentionally modelled on the MPL in order to make it GPL- incompatible. For everyone who wants to hear this first-hand, we have video from that talk available.
Here is the FSF position about the CDDL. This thread contains statements on the issue made by Debian people; for more context also see the other mails in that thread. In short -- the CDDL has extra restrictions, which the GPL does not allow. Jörg has a different opinion about this and has repeatedly stated that the CDDL is not incompatible, interpreting a facial expression in the above-mentioned video, calling us liars and generally appearing unwilling to consider our concerns (he never replied to the parts where we explained why it is incompatible). As he has basically ignored what we have said, we have no choice but to fork. While the CDDL *may* be a free license, we never questioned if it is free or not, as it is not our place to decide this as the Debian cdrtools maintainers. However, having been approved by OSI doesn't mean it's ok for any usage, as Jörg unfortunately seems to assume. There are several OSI-approved licenses that are GPL-incompatible and CDDL is one of them. That is and always was our point.
For our fork we used the last GPL-licensed version of the program code and killed the incompatibly licensed build system. It is now replaced by a cmake system, and the whole source we distribute should be free of other incompatibilities, as to the best of our current knowledge.
Anyone who wants to help with this fork, particularly developers of other distributions, is welcome to join our efforts. You can contact us on IRC, server irc.oftc.net, channel #debburn, or via mail at debburn-devel@lists.alioth.debian.org. Here is our svn repository.
In the past, we, the Debian maintainers of cdrtools, had a good and mutually cooperative relationship with Jörg Schilling. He even commented on Debian bug reports, which is one of the best things an upstream maintainer can do. Naturally, there were occasionally disagreements, but this is normal.
Unfortunately Sun then developed the CDDL and Jörg Schilling released parts of recent versions of cdrtools under this license. The CDDL is incompatible with the GPL. The FSF itself says that this is the case as do people who helped draft the CDDL. One current and one former Sun employee visited the annual Debian conference in Mexico in 2006. Danese Cooper clearly stated there that the CDDL was intentionally modelled on the MPL in order to make it GPL- incompatible. For everyone who wants to hear this first-hand, we have video from that talk available.
Here is the FSF position about the CDDL. This thread contains statements on the issue made by Debian people; for more context also see the other mails in that thread. In short -- the CDDL has extra restrictions, which the GPL does not allow. Jörg has a different opinion about this and has repeatedly stated that the CDDL is not incompatible, interpreting a facial expression in the above-mentioned video, calling us liars and generally appearing unwilling to consider our concerns (he never replied to the parts where we explained why it is incompatible). As he has basically ignored what we have said, we have no choice but to fork. While the CDDL *may* be a free license, we never questioned if it is free or not, as it is not our place to decide this as the Debian cdrtools maintainers. However, having been approved by OSI doesn't mean it's ok for any usage, as Jörg unfortunately seems to assume. There are several OSI-approved licenses that are GPL-incompatible and CDDL is one of them. That is and always was our point.
For our fork we used the last GPL-licensed version of the program code and killed the incompatibly licensed build system. It is now replaced by a cmake system, and the whole source we distribute should be free of other incompatibilities, as to the best of our current knowledge.
Anyone who wants to help with this fork, particularly developers of other distributions, is welcome to join our efforts. You can contact us on IRC, server irc.oftc.net, channel #debburn, or via mail at debburn-devel@lists.alioth.debian.org. Here is our svn repository.
They told him to fork off.
I understand dropping his package, but kicking him? Man, I don't want to upset the Debian team.
Won't the GPLv3 be incompatible with the GPL?
Fellowship 9/11
They refer to MPL in the message and I wondered if that's that Mozilla license and if that is really incompatible with the FSF.
Yeah, I'm as old as my UID would suggest.
Because most of the thousands of OSS cd tools are merely front-ends to cdrecord.
Fascism starts when the efficiency of the government becomes more important than the rights of the people.
Anyone who kept track of Joerg Schilling, and his prominent ego, was able to clearly see the inevitable fork from quite a distance away. Schilling was another one of those types -- like the dude who was running some obscure piece of code known as xfree86 -- whose success and prominence as the author of a popular free software package went completely into his head.
No, this should not be suprising news to anyone who's been following LKML. You could've predicted this a long time ago. What is really interesting here is the revelation that Sun explicitly made CDDL intentionally incompatible with GPL. That is, what I think, the newsworthy fact, and should be a wake up call to all the Sun fan club who've been slobbering all over themselves on the account of Sun's promises of releasing Java as free software.
Reading this just underscores the fact that you just can't trust Sun, and nobody should hold their breath on account of Java.
What Danese Cooper says is wrong. I and many other members of the OpenSolaris project know for certain that SUN did not create the CDDL to be purposefully incompatible with the GPL. SUN even releases other software under the GPL and LGPL.
It is also important to note that Danese Cooper's employment with SUN ended in March of 2005 (http://blogs.sun.com/DaneseCooper/). This means that any statements made by her are not officially representative of SUN. Conspiracy theorists are free to believe what they wish.
In addition, what the maintainers have failed to mention is that they have repatedly introduced patches to the codebase that have broken or otherwise caused problems in the cdrtools codebase. They need help because they don't know how to maintain cdrtools properly.
In addition, there are currently problems with Debian's Free Software Guidelines. Notably that the project does not consistently enforce them because many rules are not explicitly written, instead each software is judged on a case-by-case interpretation making it difficult for upstream developers to comply and those interpretations themselves are not always consistent. If you want proof of this, just read the various flame wars on debian-legal, etc.
I thought that someone already forked this long ago because of problems with Joerg Schilling mucking around with the license? Read the wikipedia entry on dvdrtools. In fact, dvdrtools is already a debian package. Why did they need another fork?
I suspect that, as usual, the license issues are really just an excuse, and that the real reason is that the current maintainer of cdrtools hasn't been doing a very good job. What's up with IDE being deprecated? Why do they make us go through all that SCSI business, when at least 95% of the people who use it have no SCSI? Hopefully the debian fork will make the cdrtools better and more usable.
Not according to the FSF themselves, who list it under the heading 'The following licenses are free software licenses, but are not compatible with the GNU GPL'.
Try "legal" and "illegal". If two licenses are incompatible, then it is illegal to distribute software written under both together. So it's illegal for Debian to distribute the CDDL licensed bits of cdrecord with the GPL licensed bits. This isn't squabbling about the definition of free, it's about protecting a volunteer project with no money from being sued.
"I may not have morals, but I have standards."
I understand that, but why is it grounds for a package to be removed from Debian?
I couldn't find that in the article, is this your personal inside information, did you talk to him or are you just asuming it, as it is so easy to interpret decisions in a way that fulfills your own prejudices.
"People who are willing to sacrifice essential freedoms for security deserve neither freedom nor security."
B F
Some of us grew tired of his rantings about: ...
- why scsi emulation was better than native atapi/ide support
- why the dvd patches were unofficial, and dangerous and you should buy his dvd modifications instead.
- his insistance of clearly marking "unofficial" versions with warnings that tell you to use or buy his version
- his sections of code that were not to be modified because he was afraid of answering questions about others instable patches.
- his license change
-
cdrtools is dead. long live cdrkit.
The Debian side itself says in the message that Mr. Schilling's is the original upstream code, and that he has been very supportive of them in the past.
It almost sounds as if they wanted to dictate to him what the terms should be, and they are unhappy that he is not complying.
So, does this mean Jörg's cdrtools will go the way of XFree86 4.4+?
I can see a lot of positive things coming out of this move.
Slagborr
It's true that Jörg Schilling contributed to OSS for many years and we should all thank him for that.
However, I was pretty disappointed the day I got to his site and saw that I had to pay for cdrecord if I wished to burn... a DVD ?! For crying out loud...
This kind of event is actually hindering for the OSS community in general. During years no one needed to create a set of cd-recording tools for Linux, because... there were already Jörg Schilling's ones ! Until one day, he decides to put a lid on further enhancement of his old "free" package and creates a semi-commercial product.
Now someone will have to start almost from "scratch zero" to create/evolve the new "free" cd/dvd burning tool for GNU-based operating systems.
read the fine manual. There IS a way to set the delay to 1 second or none at all.
I think they forgot to mention all the other bullshit as well. The upstream cdrecord had many problems which lived on for a time in Debian's patched version, including;
- Not being able to burn via device node. You have to specify some cryptic sequence of numbers.
- Not being able to burn as a non-root user. WTF?
- Lot's of FUD in the program output about how you should use Solaris instead of Linux.
Any bug reports relating to this on Debian's bugtracker usually incited Joerg to long fits of counterproductive trolling. Hopefully they'll see sense and ban him from similarly messing up the cdrkit bugtracker.
cdrecord dev=1,0,0 -eject schilling.iso
It is pitch black. You are likely to be eaten by a grue.
Good for Jorg to stick to his guns. He can choose whatever license he wants to release his code under.
Of course he is. This freedom extends to releasing code that nobody else can legally use. A CDDL build system+GPL codebase isn't legal for anyone else but Jorg to distribute. More power to him.
If his new license is not compatible with Debian goals, ideals, etc, and they cant agree in a common point, ok, substitute his package for another with a more Debian-like license in that particular distribution, but is not like he became the evil lord of darkness and must be despised by everyone. We all have too much to thank to him for all what he did already.
Why didn't the author include Joerg's position on this? He didn't even provide a link to his hompage:t ml
/.) could comment on that one, since I am not a kernel hacker.
/. as well as his problems with other distros and the kernel suggest that he is. I simply don't know. But I also heard that Linux Torvalds can be a very harsh himself. Anybody want to fork the kernel because of that?
http://cdrecord.berlios.de/old/private/cdrecord.h
He also seems to have problems with Suse and RedHat as far as his homepage goes (they also include older versions) and with the Linux kernel itself. There seems to be some stuff he dislikes about the SCSI subsystem. And he seems to prefer the way Solaris handles SCSI. Maybe someone with some insight (if there are any left on
Joerg Schilling is doing excellent work. But as some others have commented there seem to be personal issues. So it is a shame that they had to use such a lame excuse to boot him. I am pretty sure the fork will go nowhere or at best use patches from Joerg Schilling proving that there never were incompatible licences.
Note that I don't argue that he might be a difficult character. Comments on
It depends how the front end is made. Is the front end linked to cdrecord?
If the front end is NOT linked (and invokes the tool via something like system()) then it doesn't matter what license the tool is written under - a GPL front end can still use it and be GPL, just as you can write non-GPL software that works on Linux.
Oolite: Elite-like game. For Mac, Linux and Windows
As FreeBSD user, I don't care much about Debian's specific decisions; but regarding cdrtools, I fully agree. The latest versions have become annoyingly FUD-dy and kind of ads for Joerg's commercial version. Fortunately, burncd (for CD) and growisofs (for DVD) work just as fine here. cdrkit will be a welcome addition to FreeBSD's ports system as well.
It's not the first time some developer's stubborn-ness resulted in a fork. That's the beauty of OSS (GPL and other OSS-compatible licenses): control freaks can't get away with it. Now let's hope some brave soul would adopt cdrkit and keep it up to date with the newest burning technology.
cpghost at Cordula's Web.
If I write code under a BSD license, anyone can use make use it. GPL, BSD, CDDL, even proprietary closed source code. Everyone has freedom.
GPL code is only free for use with other GPL code.
GPL gives you freedom the same way segregation gives you freedom -- freedom for some, not all.
Do you even lift?
These aren't the 'roids you're looking for.
Sure the VM will be free software, it just won't be GPL compatable. So you'll never be able to use GPL code in the VM and you'll never be able to use VM code in anything licensed with the GPL.
There are free licenses that are not compatible with the GPL.
Back in the 1980s, the SCSI command protocol and the old-style SCSI bus were a matched pair. Devices had ID numbers that you could set with jumpers. Devices didn't move around. There was no hot-plug or plug-and-play.
/dev are now set by the user. UDEV matches various things (serial number, manufacturer, location, etc.) to identify the device. Device numbers are dynamic and essentially random. The names are stable. Normal apps open devices by name.
/dev/sg* devices, which are intended for screwball devices that don't have normal drivers. It is similar to a modem program bypassing the /dev/tty* devices by calling iopl() and then directly controlling the hardware.
Now we run the SCSI protocol over USB, FireWire, SerialATA, TCP/IP, and numerous other transports. You can't address all the devices on the Internet with a 3-bit number. Devices come and go. If you plug in a CD burner, it usually shouldn't matter which USB port you use.
The Linux solution is UDEV. We can also use D-BUS and HAL. Device names in
Joerg wants to use an obsolete backdoor. He doesn't use the normal device names or the normal CD/DVD driver. He uses the
Suppose you have two USB burners. If you yank out your USB cable and then put it back, the device numbers may change. The device names can remain the same, thanks to UDEV. Joerg's defective program will be unaware of this. It will just use the wrong burner.
Reread the parent. He said that a project that has both code licensed only under the GPL and code only licensed with {a license incompatible with the GPL} cannot be in Debian, because it would be illegal to distribute.
This isn't about putting Apache and GNU C in the same distribution. It's about putting filemanager.c and documentview.c in the same binary when filemanager.c is licensed under the XGL, and documentview.c is licensed under the XGL-incompatible YGL. That's the core of the problem here.
You are not alone. This is not normal. None of this is normal.
Once a piece of code has been released under a license (such as the GPL), you cannot retroactively change that license (ie tell people they can no longer distribute it under that license)
There are a number of copyright holders; see http://svn.debian.org/wsvn/debburn/nonameyet/trunk /debian/copyright?op=file&rev=0&sc=0
It works like this: The CDDL is incompatible with the GPL. Schilling doesn't want to believe it is, but both the CDDL and GPL writers (and anyone with half a brain) say otherwise. So while he's perfectly within his rights to distribute source code that combines CDDL & GPL code (as he is doing now), as soon as you build that source code and distribute the result (as any binary distribution does), you've just violated the GPL's 'no additional restrictions' clause.
While Debian has the balls to do this, Gentoo already had a GPLed fork of cdrtools available, and TOOK IT AWAY just because a new version of cdrtools came out with a few new features.
The forked code was GPL'd, you cannot revoke GPL once it's given. Jorg has no say in how his GPL'd code is used, modified or distributed provided it is in accordance with the GPL version with which it was released.
NZ Electronics Enthusiasts: Check out my Trade Me Listings
Have you ever read an email by Mr Schilling? Try this thread on lkml, and tell me who is being the most annoying. He drags himself through the mud by alienating people with his attitude.
IANAL nor a Debian Developer, but as a Debian user I seem to remember that Debian has a non-free "tree" of packages, since "main" has to be DFSG-compatible. In fact, no package can be in main if its dependencies are non-DFSG. The DFSG's restrictions reads in many ways like cliff notes of someone reading the GPL(the GPL seems longer and more legalese), and explicitly approve the artistic, BSD and GPL licenses. I'm thinking the CDDL isn't just GPL-incompatible, but is also violates the DFSG, but again IANAL(although I read the thread linked from the article and they seem to agree with me). Back on your question about how licensing would make things incompatible, it only matters when, like Debian, you distribute the whole thing, with a charter and/or some agreement that all the licenses meet a specific set of criteria. It wouldn't stop say, Novell or Mandriva, which, I believe, operate under no licensing restrictions other than the licenses themselves.
The interaction of the GPL, MIT, and BSD licenses is well understood and works well.
There's no problem at all linking GPL software with libraries of either. Same goes with the apache license and perl's artistic license.
Sun's license isn't GPL-friendly, and even if there's a question about it, debian needs to find a way around it. This is the way debian works - it's all in the social contract. It's a pain sometimes, but there's distros out there who don't worry so much about licensing issues you can use if you're concerned.
Those who can't do, teach. Those who can't teach either, do tech support.
Out of general interest, I read this thread. It starts out with a guy complaining that Joerg didn't respond to email. When Joerg appears, he states that he did respond to the email, the next day. Then all hell breaks loose and everybody is immediately on Joerg's case. There's obviously a lot of historical resentment here, but this particular discussion doesn't really paint Joerg in a bad light.
ABOUT!!!
EFFFING!!!
TIME!!!
I have DESPISED this man's code since the day I saw it. His BONEHEADED insistence on doing things the Solaris way in Linux, his apparent INABILITY to use a standard build system, and the INSUFFERABLE ARROGANCE he displays through absolutely everything he does are completely INFURIATING.
Think I'm spewing flamebait? Nonsense. Read this bug report about cdrtools. He starts by insisting his misinterpretation of the GPL is correct, goes on to threaten defamation(slander) lawsuits in german courts against Debian, and finishes up calling most the people in the discussion thread "convinced liars". The man is unusable as an open source contributor, and I am ecstatic that more people actually realize this now.
Just another "DOJ fascist authoritarian totalitarian bootlicker" -- Zeio
what? if your project is committed to remaining GPL-compatible and a contributor relicenses code that's fundamentally incompatible with that, what do you expect them to do?
debian did the right thing, in a straight-forward and even gracious way.
pr0n - keeping monitor glass spotless since 1981.
Has anyone here ever tried to buy a ProDVD license? I have, on behalf of a former employer, and the result (by email) was always no reply. Every year or so our production system would stop working until someone realized that another "free" key had gone bad.
So not only does Jörg keep his software non-free - he doesn't take money for it either. I concluded a long time ago that his thought processes are not standard issue.
Sure. He can do that. He has the right.
That's a very irritating decision though, especially when he refuses patches to add the missing feature.
It's Debian's right to decide Joerg can go to Hell.
I mean, I of course see the point of removing Jörg Schilling from the equation, but the guys from ark linux have already made a clean fork a few months ago called dvdrtools ( http://www.arklinux.org/projects/dvdrtools ) ( server seems to be down at the moment ).
Malike Bamiyi wanted my assistance.
You should read the whole thread.
Any criticism on how eg USB doesn't follow Joergs preferred namingscheme goes unanswered somehow.
I'm only a simple user, but even in my experience the dev=h,b,t,l way to address a burner is flawed. Anyone can reproduce it with 1 usb burner and a couple of usb drives or simply 1 firewire disk (which will simply increase the hostid each time you unplug/plug it (atleast mine does)).
He had some credits for bringing cd recording to Linux, but maybe he should simply abandon Linux and concentrate on his beloved Solaris.
Once a piece of code has been released under a license (such as the GPL), you cannot retroactively change that license (ie tell people they can no longer distribute it under that license)
That is not entirely correct. You can legally revoke a license at any time. "License" is just a legal term for "permission" or "consent", and you can withdraw permission and consent, and so can withdraw a license. Nevertheless, if you try enforcing that revocation in a court you are likely to run into issues of estoppel. In simple terms, if somebody has relied on the license in a way that would make it unconscionable for you to withdraw it, the court will hold you to the terms even though you may have revoked it.
With GPL software, where somebody else has relied on the license and produced non-trivially derivative works (or even non-derivative works that depend on the GPL software) then withdrawing the license would be unconscionable because they have expended significant effort (capital expenditure) in reliance on the license which is lost if the license to the original software is revoked. It may also be that if other people have refrained from developing equivalent software because of the existence of this particular GPL software, then it would be taken as unconscionable to withraw the license, at least until such time as equivalent software can be produced.
On the other hand, to use an extreme example, say you have produced something and released it under the GPL, but nobody has used it. You could revoke the GPL on that software at any time. You could also revoke the GPL at any time if there is a readily available substitute provided nobody has produced any derivative work.
While it is quite common to say that you cannot revoke the GPL on a piece of software once released, this is not literally true. While in many cases this will be the situation for all practical purposes, the general rule is more complex, and in the right circumstances it is possible to revoke it.
Ah, but you see, that is not the fault of cdrtools! Oh no! Rather, it is a flaw in the (unmaintained) Linux kernel, that should have simply adopted the design of Solaris' SCSI subsystem that has been in use since the year 1981!
If you continue to experience problems then it is recommended that you upgrade to Solaris or Linux 2.4.
I think this is pretty much the file you forgot:
/ 4/en/os/i386/SRPMS/sg3_utils-1.06-3.src.rpm
http://ftp.redhat.com/pub/redhat/linux/enterprise
That is an SRPM. Red Hat doesn't seem to provide binary RPM files
for ES4. You should have an rpmbuild command that will build that
into a regular rpm. The rpm command itself used to be able to build
from source; probably the ability still exists in RHEL ES4.
Debian certainly provides sg_scan.
As for ifconfig: that is kind of obsolete now. It's a compatibility
hack that uses a sort of BSD emulation layer. Getting to know the
more-powerful native tools (the ip command) would be a good idea.
He'd be in the clear if cdrecord were 100% his own work. If that were the case, then one might reasonably argue that he has implicitly granted an exception to the GPL and/or CDDL.
Problem for Joerg: he has included GPL work from other people. This puts Joerg in violation of the GPL.
"his thought processes are not standard issue"
Oh my. That is perfect.
I was on the fence as to whether to hold the parent's opinion (which is that some projects have NIH issues that come out like this), or the general opinion that Joerg Schilling is seriously messed up. So I did some research. Read through these threads and I believe you will come to a point of view that is balanced and based on primary sources of evidence. You can also read my commentary, however it would be considered a secondary source of evidence.
0 113.html
http://lists.debian.org/debian-devel/2006/08/msg0
There is a LOT of material here, I'll break down my impressions:
- Some instances of misunderstood word usage (ie CDDL is no longer acceptable to the Debian project in what appears to be a new bylaw or whatever they use)
- as a result Joerg Schilling accuses the project of being 'untrustworthy' and 'suspect.' IMO Far too strong terms to use for what could at worst be described as inconsistency. And I'm stretching that definition.
- REFUSED a request to move legal discussion to more appropriate mailing lists and claimed that personal accusations/attacks were made upon him on debian-devel and pointed out his own feelings that he should defend himself on debian-devel (This seems like such a breach of decorum after a civil if difficult debate/raging argument)
- Interpreting the GPL Preamble as word of law (after failing several other dubious GPL interpretations and basically accusing the FSF GPL FAQ maintainers of not knowing how to do their job)
- Having finally been pegged to a request for a name change in the event of a fork, tries to lay claim to the name 'dvdrecord' despite having dubious ability to claim ownership of a generic trademark like 'cdrecord' in the first place.
On the whole debian-devel participants displayed an AMAZING sense of decorum and civility in the face of nonsensical diatribes and difficult debate. Props to them.
There is a more technical debate dated around February on LKML regarding libsgc and cdrecord. (no link simply because I'm having trouble finding the head of the discussion, search LKML Jorg to find.) Here it appears that Joerg Schilling simply appears to be unwilling to compromise functionality and code in order to make his software work properly (or sanely) under Linux. This is closely tied into a unique view on how to make his code cross-platform and the fact that libsgc is meant to integrate with a far greater generalization of SCSI components than just CD drivers.
I don't understand the technical intricacies but it appears that over time the SCSI and IDE interface has changed dramatically. As a result he believes the kernel should change to accomodate his software. That wasn't received very well at which point it might have been appropriate to chalk it up to simple disagreement and walk away, however it degenerated into a variety of other semi-related discussions that were far too personal. (Keep an eye open for 'smake' and the 'Schily Makefile System,' I kid you not.)
"That is not entirely correct. You can legally revoke a license at any time."
...and...
Not this one, because the license terms themselves:
"2.b) ou must cause any work that you distribute or publish, that in whole or in part contains or is derived from the Program or any part thereof, to be licensed as a whole at no charge to all third parties under the terms of this License.
6. Each time you redistribute the Program (or any work based on the Program), the recipient automatically receives a license from the original licensor to copy, distribute or modify the Program subject to these terms and conditions."
Given them both together it means that while it might be within your powers to revoke the license to those you directly distributed to (since it's a matter about *you* and someone else, and even then, as you properly stated, it will be quite difficult to convince a judge you can break the confidence of your licensees without a really strong reason), you can't deal on something that it is not your bussiness, that is, the deal among "second tier" redistributors and their receptors. So you, as most, can avoid people that recieved copies directly from you to further redistribute, but you won't be able to avoid redistribution from people that didn't get the code from you, much less those that got the code neither from you nor you direct "clients".