Morosoph's Journal: Linus is wrong over the GPL version 3

Torvalds reveals how technical competence does not imply philosophical depth.

Here is the devastating evidence: a letter to The Register.

Now, I know that The Register isn't exactly the most highbrow publication on the planet, but let us analyse Linus's reply to The Register's earlier article:

Since you seem to be following the kernel mailing list, you could have picked a better email to comment on.

Okay, The Register was seeking to put Linus in a bad light; they could have been fairer to him.

Anyway, look for the one that talks about "reciprocity" to get an idea for why I like the GPLv2, and not the GPLv3 (I don't know how you read the mailing list, but you can search for it at least on lkml.org).

Reciprocity optimises a two-entity system; writing software, however, is a manifold system. The virtue of the GPL is that it spills onwards, and requires that others who want to use its bounty have to reciprocate with the GPL's conception of freedom; not with a particular individual, or collection of individuals. The future developer or user is protected by the GPL. The GPL creates a freedom akin to the free market, and as such protects future transferability, and future trade.

In a manifold system, "reciprocity" is often cartelisation. Higher values turn one against the immediate group for the sake of the larger group. I imagine that Linus is thinking that since he'd like hardware manufactorers to respect programmers' freedoms, so he should act to protect theirs.

It's not about "freedom". It's about "fairness".

Which, btw, is a lot more fundamental concept. "Do unto others.." and all that, you know.

And here Linus betrays where he's really coming from. This is pure Scandinavia. And here he also reveals a lack of philosophical depth.

I know from hard experience that "freedom" is often seen as being synonymous with greed, but it doesn't take long to see why this is mistaken: freedom is essentially the flipside of trust, but trust is a complex beast when one is dealing with something as complex as society. We (by and large) accept laws in order to allow trust to develop. Is other words, we selectively distrust in order to raise the overall level of trust. The GPL functions much in the same way as law does. The BSD licence is more trusting, mostly because programmers who write under BSD licences want it that way. The GPL, however, creates trust further down the line, and trust from those who might wish to contribute to the body of the GPLed work. The GPL is deliberately unfair to the closed shop, for much the same reason that green taxes are deliberately unfair to polluting entities; in both cases, they are seeking to change behaviour. The GPL aims to maximise freedom, and green taxes aim to maximise the difference between the value of 'goods' and 'bads'.

Fairness is just about the worst-defined concept in common use. The golden rule fails as soon as you realise that people want different things (or else there wouldn't be trade), and just about everyone defines fairness as the perfection of their chosen strain of politics (equal opportunity, equal outcome, equal consideration, trade without compulsion, ...), so that to use the word is essentially to have said nothing over and above "it doesn't suit my politics". When you go about defining exactly what should be made fair, then you have something to talk about.

It is my opinion that any definition of fair which justifies the GPL version 2 cannot help but justify version 3, and even prefer it. I understand that the GPL version 2 may be a good compromise for many individuals and entities, but that is exactly what it is: a compromise.

Note to RingDev: I'm sorry, I was (mostly) away from the computer for a few days, so I lost the chance to follow up on your reply, so I posted a delayed response in my last JE.

I'd go further

[Concern]

Okay, The Register was seeking to put Linus in a bad light; they could have been fairer to him.

I don't agree. I think they were exceptionally balanced and thus unjustifiably charitable towards the guy.

As far as I can tell, Torvalds has no ability to relicense Linux, since unlike at the FSF, copyrights were never assigned to some central authority (or at least not comprehensively). Every single person who ever contributed a line would have to independently agree to a license change. In other words, later tha

Re:I'd go further

[Morosoph]

I fault them for not going harder on the guy's obvious mistakes.

This much I concur with. I prefer to avoid focusing upon motivation, though, as it's very easy to get wrong. Also, the press end up looking bad, and less subtle readers pick up a much cruder message than the one that is outwardly stated. The way that you feed intelligent discrimination is in part by starving cruder discrimination.

I agree WRT the black hole. There is a serious issue about possible unintended consequences of the licence

Next?

[peacefinder]

As I understand it (which is dimly) there's a fair bit of code in linux that is licensed GPL2 instead of "GPL2 or later". If so, relicensing would present a huge headache for Linus that I'm sure he's not eager to take on.

In any case, even if Linus himself declines GPL3, can't someone else make a derivitave work (by replacing linux's GPL2-only bits, for instance) and publish it themselves under GPL3? Wouldn't that be sufficient?

Re:Next?

[Allen Zadr]

No.

The GPL is a license. The license states that the software must always follow the license, in perpetuity. Linus owns copyrights to most of Linux, and Linus specifically licenses Linux under GPL v. 2 (not any GPL). Thus Linus extends to you, the right to use and derive the software under GPL v. 2. So, until Linus chooses otherwise, you may not change the licensing terms based on what Richard Stallman has to say on the matter.

Yet, all of this is basically moot. A GPL v 3 version of the Kernel w

Re:Next?

[Concern]

Ah... this makes no sense. GPLv3 does not say "you cannot use crypto." There is no sane reason to remove crypto code from GPLv3 work. It just says if you're going to take some GPLv3 code and make it _primarily_ for hardware with a chain of trust, where some signing key is _required_ to use it (that is, not _optional_, as for instance Linus or Redhat's keys are), then you need to disclose the key... or, just not use the GPLv3 work.

This is just about preventing the use of hardware as a loophole to take GPL co

Clarification

[Allen Zadr]

I've read it carefully, and it's not worded that way. It specifically says, "It also includes any decryption codes necessary to access or unseal the work's output." and "No covered work constitutes an effective technological protection measure..."

Yep, that would include PGP. Of course, what does DRM do that PGP doesn't do just as effectively. What's technically different from a Kerberos use ticket to an Apple DRM AAC file?

The wording is very specific, and repeatitive (Pramble,

Re:Clarification

[Concern]

Just listen to yourself.

Do you really think the GPLv3 is incompatible with crypto? Do you imagine that they would do that? It's absurd.

This is just inflamatory nonsense on its face - sadly, encouraged by Linus.

The sentence you quote does not mean what you think it means. Read the one after it: "Notwithstanding this, a code need not be included in cases where use of the work normally implies the user already has it." This does not apply to crypto APIs, only to systems where there's some particular secret key

Re:Clarification

[Allen Zadr]

It's not that hard to distinguish between prison platforms and general and voluntary uses of encryption.

Really? Personally, I don't see any difference at all, at least not from a technical standpoint. Both "DRM" and "Encryption" start with a block of data, which is scrambled and only supposed to be opened by the person who has the key. In both cases, multiple users may be able to unlock the code.

I read the whole thing, multiple times, and unlike you, I'm not willing to go on blind faith that it do