Subject: Re: brainstorming
From: Brian Behlendorf <>
Date: Sat, 15 Jan 2005 23:57:57 -0800 (PST)

Ben, it can't be mere coincidence that you start this thread on the same 
day that IBM makes its announcement of its "patents pledge":

And see the corresponding discussions and articles linked from slashdot,, etc. I'm not sure why IBM decided to call it a "legally-binding 
pledge" rather than a license, but I'll call it a license for shorthand.

It's a very good thing to see happen.  It's a defensive move in what 
people are worried would be the next SCO-like attack on Linux or other 
open source projects.  While the press releases and articles trumpted the 
"value" such ideas might bring to the OS community if implemented, the 
bigger value in publishing this is being able to raise the costs for 
anyone considering initiating a lawsuit against an Open Source project - 
because if you do, your rights to use *any* OS-licensed software that 
implements those patents is terminated.

Even though this license can apply to any OS-licensed software, it
has a different effect on derivative works than the OS license itself 
might have, simply because the license doesn't also extend to derivative 
works that are not OS-licensed.  For example: if a corporation were to 
download an IBM-patent-encumbered Apache codebase from, 
incorporate that codebase into their proprietary work, and then 
redistribute the resulting work under a non-OS-license, then the IBM 
patent license would not extend to this new derivative work - even though, 
by the terms of the Apache license, releasing a non-OS-licensed derivative 
work is perfectly fine.  This is not *quite* like the GPL since it appears 
that the corporation could probably find a way to redistribute the 
patent-encumbered code under an OS license without releasing the whole 
derivative work; so I won't use the term "viral", but it's at least 
an additional complication.

If, if, IBM were to say that the patent license applies to any 
OS-licensed-work, as well as any derivative of that work (so long as the 
patent being licensed was originally in the OS-licensed-work) that would 
solve the problem.  It would still mean that licensees would have to 
refrain from patent enforcement against OS projects, something the Apache 
license does not require except in a very limited circumstance.  But... I 
think that's a reasonable additional restriction.

Unless such a change is made, though, anyone writing BSD or 
Apache-licensed code would probably do well to *not* examine the list of 
patents that IBM has granted, for that would (I contend) more or less 
compell them to note any patents that may apply, and that would in essence 
constitute a major additional restriction above those licenses.  GPL 
authors would never need to worry; nor would authors who publish under 
weak copyleft licenses like the MPL or CPL.

Making sure that patent lawsuits are a mutual-assured-desctruction 
scenario seems more useful (to me) than compelling Open Source 
redistribution of patent-encumbered works.