Linux 6.7 Set To Drop Support For Itanium IA-64

[mercster]

Yeah I feel ya. But now most Linux devs are corporate-powered so they can’t poopoo the cashman.

Yeah. And look, I'm not some GNU fanatic Stallman-type. It's an overall plus that big companies got behind Linux and it has corporate support and all that. It kinda has to, there's no way pure volunteers could keep it where it needs to be to be as powerful as it is in today's ecosystems. But it's different... before it was some idealistic "Let's support everything and give everyone freedom!" vibe, and now.... there are costs involved. It is just the way of things, I suppose.

That's why I said, I'm not expecting the kernel devs (to whom I freely admit I have nothing to offer) to carry this on their back but, there must be some kind of... I dunno. Just kinda let it limp along. I have a feeling they look at lots of automatic unit tests and get aggravated when something in that tree gives them a headache. I dunno. I'm guessing.

[jabl]

For what it's worth, I agree. I don't see the same push to remove PA or Alpha, and those are a hell of a lot older and slower than Itanium is.

PA and Alpha still have active maintainers, Itanium doesn't. Simple as that really. Maybe there's a better supply of leftover PA and Alpha HW that hobbyists can pick up on the cheap?

(Some years ago there was a bit of de-Alphaization in the core kernel going on. Alpha has the craziest least strict memory consistency model you can imagine, and many of the core kernel synchronization primitives were made with Alpha in mind as the lowest common denominator. Most of this has now been removed, making the core synchronization primitives a bit simpler, at the cost of Alpha being more tightly synchronized than it needs to.)

[mdedetrich]

Maybe it's just because I was using Linux pretty early, but it always feels weird when the kernel removes support for some architecture. Back in the mid-90s doing 'make config', it was always impressive to see all the architectures, devices, and protocols supported... maybe only a small number of people need that, but it was a badge of honor for the Linux kernel, that it was a sort of Swiss army knife that would suit many different purposes.

Nowadays the world is different, I understand that... and I do sympathize with the burden "dead" branches in the kernel might cause to developers and maintainers in other parts of the tree. Still... seems like there could be some kind of alternative to either keeping it in the main tree, and removing it. I dunno... maybe not. It's just an emotional reaction on my part, I reckon.

(My mind is drawn to AX.25 packet radio in the kernel, for instance. Is there a whole lot of people using it these days? Probably not... but someone out there might be. Support for all these exotic older standards is something I feel like should somehow be preserved. I don't even know if AX.25 is still in the kernel...)

(Note 2: To say "IA-64 users are all using HP-UX and OpenVMS anyway"... again seems strange to me. Back in the good ol' days, someone might have remarked "But why should their only choices be closed? Doesn't all hardware deserve a free and open alternative?" Again, I know I'm probably just being romantic.)

I guess you can say is what has changed is that Linux is less of a hobby/toy project now, which means if they say they support something they really mean it and if they can't support something (i.e. they don't even have access to physical hardware to see if kernel compiles etc etc) then they will consider removing it.

Even GCC is slated to remove IA64 at some point, so the kernel doesn't really have a choice here unless they want to stick to an old GCC version.

[mercster]

I guess you can say is what has changed is that Linux is less of a hobby/toy project now, which means if they say they support something they really mean it and if they can't support something (i.e. they don't even have access to physical hardware to see if kernel compiles etc etc) then they will consider removing it.

Even GCC is slated to remove IA64 at some point, so the kernel doesn't really have a choice here unless they want to stick to an old GCC version.

Yeah, I understand that. Again, this is more an emotional response than anything... the world's in a different place. I remember back then, interested hackers would KILL to get some piece of rare/expensive hardware so they could port Linux. It didn't matter that it was economically justifiable. People would donate to a skilled hacker, so he could buy... oh I dunno, some weird expensive enterprise-level architecture, so that they could get to work porting Linux to it. Now, I guess hardware is much less heterogenous and all the diversity is in software.

Like I said, no judgement here. Just... sad to me. Crying over spilled milk that noone's missing, probably.

[Eirikr1848]

Yeah, I understand that. Again, this is more an emotional response than anything... the world's in a different place. I remember back then, interested hackers would KILL to get some piece of rare/expensive hardware so they could port Linux. It didn't matter that it was economically justifiable. People would donate to a skilled hacker, so he could buy... oh I dunno, some weird expensive enterprise-level architecture, so that they could get to work porting Linux to it. Now, I guess hardware is much less heterogenous and all the diversity is in software.

Like I said, no judgement here. Just... sad to me. Crying over spilled milk that noone's missing, probably.

“LINUX ON YOUR TOASTER”. Now even the Revolution Smart Toaster doesn’t have a working Linux build…

We have now moved from “Linux on everything” to some grumpy goose being all like “FiNaLLy RiD of CrUfTy CruFt”.

[Eirikr1848]

Yeah. And look, I'm not some GNU fanatic Stallman-type. It's an overall plus that big companies got behind Linux and it has corporate support and all that. It kinda has to, there's no way pure volunteers could keep it where it needs to be to be as powerful as it is in today's ecosystems. But it's different... before it was some idealistic "Let's support everything and give everyone freedom!" vibe, and now.... there are costs involved. It is just the way of things, I suppose.

That's why I said, I'm not expecting the kernel devs (to whom I freely admit I have nothing to offer) to carry this on their back but, there must be some kind of... I dunno. Just kinda let it limp along. I have a feeling they look at lots of automatic unit tests and get aggravated when something in that tree gives them a headache. I dunno. I'm guessing.

Oh also. As someone who spent way too much on Itanium hardware in 2021 for personal use trying to get something that was non-x86 to use with Radeon VII cards for compute I like… super regret my choices. Shoulda gone Epyc or even POWER. Shoulda gone Nvidia. Alas.

So. I would be the first (probz only) to enjoy that limp-along support

[mercster]

“LINUX ON YOUR TOASTER”. Now even the Revolution Smart Toaster doesn’t have a working Linux build…

We have now moved from “Linux on everything” to some grumpy goose being all like “FiNaLLy RiD of CrUfTy CruFt”.

ROFL... exactly. "Get rid of this useless crap!" 😂 And I'm in no position to complain. I only lament.

[uxmkt]

Makes me wonder if it would have been cheaper for (HP) have just [..] mutually agree to cancel the contract. I figure the few remaining HP customers who use it aren't worth maintaining the architecture [for], even if HP were to get sued.

IANAL but I don't think you can really have "uncancelable" contract. At worst, it's cancelable but the cancellation fee is just set so high that cancellation is not practical so no one will cancel. (And then one can also thinking about splitting the company and and possibly letting the spinoff go bankrupt in short order. There's even https://en.wikipedia.org/wiki/Texas_two-step_bankruptcy I found...)

[oiaohm]

IANAL but I don't think you can really have "uncancelable" contract. At worst, it's cancelable but the cancellation fee is just set so high that cancellation is not practical so no one will cancel. (But then, there are things like https://en.wikipedia.org/wiki/Texas_two-step_bankruptcy ...)

There is such a thing as what classed as "uncancelable" contract. In the Intel-HP case is HP put up the money for the fab that was being used to build the chips. Intel attempts to cancel contract the terms is Intel has to give up the IP to make the chip and the Fab to make it and the income the Fab has generated on other products in other words ouch.

Another name for these "uncancelable" contract is "secured creditor contract". These secured creditor contracts process automatically as soon as you apply for bankruptcy with the secured creditors getting everything their contracts says they will one way or the other before any other creditor gets anything. There is no way out of a secured creditor contract that is valid legally other than completed. Secured creditor contract cannot be canceled they can only be completed. Of course all parties in a secure creditor contract could agree to declare the contract completed without everything in the secured creditor contract being done.

[rene]

I guess there is at least one Linux distribution we can count on continuing IA-64 support: https://t2sde.org ;-)