Hall: IBM, Red Hat and Free Software: An old maddog’s view

The i386 PC won, thanks to the fortunate events, and size, and execution, of IBM, Intel and Compaq from the 8086 PC on. But had that not happened, some other combination of manufacturers would have had mass market 32-bit, affordable (if not super cheap), MMU equipped computers.

Why imagine it, when Acorn, Amiga, Atari, etc., existed and had such machines on the market (the Amiga 3000 with the MMU equipped 68030, the Archimedes). That they didn't succeed against the PC is not an argument that the market would not have existed without the PC. Bizarre.

The entire business model of game consoles has (at least historically) been to sell moderately powerful computers at slightly below cost, and make up the losses in game sales. If you can't sell (enough) games because you can't (sufficiently) control the OS, then you have to raise the price of the console above the price of a comparable gaming PC (or some competing console), and then consumers notice the price is too high and jump ship. That does not apply to generic commodity computers, which are normally sold above cost, the manufacturer does not care what happens after the consumer buys one, and there is (for the most part) nowhere else for consumers to go if the price is too high (aside from other manufacturers of the same product).

It sounds like your claim is that a similar business model (sell lots of software to subsidize your underpriced hardware) would have developed for commodity computers, but I find that difficult to believe. The whole point of games is that they are, to some extent, consumable. You play a game, and then, eventually, you stop playing it and go play something else instead. You may, eventually, return to an old game, but you're still going to want to play something new every now and then, so you keep buying new games. This is not how generic productivity software works - users buy the software they want, and then it may be many years between paid updates, particularly in the time period we're talking about (i.e. long before the subscription model became standardized). You simply don't have enough of a steady income stream to pay for all the hardware.

One thing is certain though, when *many* companies _have already_ been working on building ever more capable, ever cheaper, ever more commodity computers; as most of society (even outside the actual people in the industry) are agreed computers are going to revolutionise things; it was _inevitable_ we were going to get *exactly the thing those many companies were already competing to provide*: A capable, affordable computer suitable for hobbyists to tinker on and write OSes that were never going to amount to anything.

tl;dr: Had Intel, IBM, Compaq - and the following army of cloners - not won dominance, Linus would have been hacking on an Amiga 3000, or an Atari ST, or an Alpha, or... etc. And we would still today have computers everywhere (from your pocket to your car to servers, to the moon) running Linux. Regardless of architecture.

And gosh, Linus *was* hacking on an Alpha by the mid-90s. ;) Maddog didn't mention he arranged that. (And the little Alpha 150 workstation DEC made was fairly cheap - almost affordable; I think Slashdot began running on one of those!).

That's not how any of this works. Estoppel is an equitable right that attaches under a wide variety of circumstances. I am skeptical that it attaches under *these* circumstances in particular (at a minimum, you would need to produce a written record of the FSF and/or GNU explicitly and unambiguously endorsing a particular interpretation of the GPL, and even then it would likely only bind those parties and not e.g. Linus), but to assert that it never applies to copyright at all goes way too far. For example, a promissory estoppel might be styled as an "implied verbal license" or something like that, but the effect is the same: The plaintiff loses their case.

Tortious interference, on the other hand, is a matter of contract law. Since most copyright licenses are contracts, including the GPL in most jurisdictions,[1] it follows that tortious interference with a copyright license is possible, but again, I'm not seeing how it applies here. Vaguetweeting about how you don't like Red Hat's interpretation of the GPL can hardly rise to the level of tortious interference. At a minimum, you'd have to explicitly solicit RH's customers to breach their support agreements, and I am not aware of the FSF (or anyone else) actually doing that.

[1]: https://lwn.net/Articles/747563/

Uh, you're the one saying that RH chosing to terminate the agreement constitutes a "further restriction" under the GPL.

> So it's Ok if RH adds any kind of further restriction?

Of course not. But RH isn't adding restrictions to the GPL rights; these restrictions are on their service agreement. To quote RH:

"This Agreement establishes the rights and obligations associated with Subscription Services and is not intended to limit your rights to software code under the terms of an open source license."

> The customer receives binaries, sources, but is only permitted to distribute sources, if the customer pays RH $1 million.

Please, there's no need to discuss a ridiculous hypothetical here, but I'll humor you.

> That would be GPL compliant?

It entirely depends on the agreement the two parties signed. Is this software the vendor wrote themselves? They can put it under whatever terms they want. Is this software modified just for that customer, for a fee? Then it's work-for-hire and the parties can structure that fee however they want. (I will note that it's common for there to be a "buyout clause to obtain full/perpetual rights" in commercial agreements)

If the "Vendor" is taking third-party software and redistributing it (modified or otherwise) then they can't impose any restrictions beyond what their upstream license (ie GPL) allows. So your hypothetical would be nominally a GPL violation -- though according to the GPL the customer would be free to just strip off those terms and carry on.

However, there's nothing to stop the vendor from just suing the customer for breach of contract and having a decent chance of winning -- You can waive all sorts of fundamental rights in a contract, as long as there is no coercion involved.

Seriously. I don't understand why folks are arguing about this, given that even the FSF has (perhaps reluctantly) blessed this sort of arrangement since before Linux ever existed, and instead of addressing this "problem" in the GPLv3 (finalized seven years after the first RHEL release!) or even the GPLv2 (Released in 1991, by which point Cygnus had been doing the same thing with GCC for at least four years) the FSF instead strengthened the language making it explicit that access to any form of support or future updates falls outside the scope of the GPL.

So, no, what RH is doing both falls in line with both the letter, and the spirit, of the GPL.

No, that would be a specific restriction on a GPL granted right. A right not granted by RH, a right (mostly, RH contributes to many packages) RH isn't able to take away.

The rights RH may take away are those that were born of the subscription agreement.