Gentoo Linux to drop Itanium support as Funtoo fork enters 'Hobby Mode'

News is bubbling up both from the Gentoo project and its successor, the tellingly named "Funtoo" – what Gentoo founder Daniel Robbins did next. The source-based Gentoo Linux distribution, which still supports a wide range of CPU architectures, will soon support one less. The project leaders announced that it is removing …

[karlkarl]

> Funtoo fork enters 'Hobby Mode'.

Was Funtoo not always hobby mode? There's nothing wrong with that, it didn't need to claim otherwise.

>> Welcome! Funtoo Linux (distrowatch) is a community-developed Linux meta-distribution and evolution of Gentoo Linux.

Another example where the "evolution" doesn't quite pan out. The Lindy effect is real!

[Liam Proven]

[Author here]

TBH, I read a number of documents purporting to explain how and more importantly why Funtoo differed from Gentoo, but I was not able to extract enough meaningful info to summarise it in the article.

I remember meeting with a group of Gentoo users and developers at a Linux conf in London about 20 years ago, and trying to explain to them that Gentoo had no answers for the sort of stuff I wanted to customise in a Linux distro. It was not possible with their tools. They were baffled. Why would anyone want a pre-SysV init, such as a BSD init? Why would anyone want to tick a box to disable dynamic linking and use static binaries? Why would anyone want a read-only root filesystem with all state banished to a specific named directory tree, which could be mounted over the network for a totally stateless client?

It's very much the same sort of experience I had meeting with some core KDE developers in 2022.

They *think* they provide vast customisability but it's only because they have very narrow, limited imaginations and are unaware that there is a box to think outside of.

[boblongii]

As a very long-time and happy Gentoo user I know exactly what you mean but when I use Ubuntu (which I must for work) it makes me shudder. Cannonical not only don't want to think outside the box, they want you to get in the box they've built and stay there.

There's a fine line in software development - not just Free or Open - you want to pay enough to get useful things done but not so much that constantly making changes to the existing software is a viable business plan.

[Liam Proven]

> Cannonical not only don't want to think outside the box, they want you to get in the box they've built and stay there.

Well, up to a limited point, yes.

I have tried a _lot_ of Unixes over the last 1/3 of a century.

In the 1990s my more knowledgeable mates told me that Debian was the way, the truth and the light. I tried. It was too damned hard. Even things like Stormix or Xandros, which helped, only helped a little and with caveats.

RHL was much too hard as well, incidentally. Worse.

So I ran Caldera OpenLinux. When that petered out, I ran SUSE.

The Ubuntu came along and over a few releases made Debian easy. Ubuntu demystified Debian, and in the end, many of the changes filtered back upstream and Debian got a lot easier too.

Fedora is from RH which has the worst case of NIH syndrome in the FOSS world and _refuses_ to learn from other vendors. Life is too short.

Devuan is Debian from people who liked it when it was more arcane and difficult. If that's your thing, go for it, but I am not masochistic enough.

Now I am eyeing up MX Linux and Alpine as candidate replacements for the huge monolith that Debuntu is metastasizing into.

[IGnatius T Foobar !]

Now end Windows

Itanium got the end it deserved. Now please rid the world of Microsoft Windows. It's far worse than Itanium.

[Liam Proven]

Re: Now end Windows

[Author here]

> Now please rid the world of Microsoft Windows.

Deep inside Windows NT is a really good OS that MS could still excavate if it had the will.

Some off the cuff brainstorming...

Remove the GUI from the kernel completely. To hell with performance overheads. Nobody cares any more. Banish it back to userland where it was in NT 3.x.

Remove networking; put it back in a module that can be removed totally, as it could in NT 3.x/4.x. Make it possible to remove it, purge all config, and reinstall it at will.

Kill WSL, which was a mistake. Make the POSIX environment mimic Linux, and make it optional. If that's too hard, make it FreeBSD or NetBSD.

Banish all traces of .NET. That was only there in case the DOJ split 'em up. Burn it in a fire; it was a bad plan. Leave only the cross-platform FOSS stuff, entirely separate.

Strip it down to a bare text-mode kernel, akin to its ancestor VMS, and maybe even drop Win32. Make Win64 the only option. That kills 90% of the malware.

Then build it back properly, all modular, all removable and purgable without killing the OS. GUI apps run only in isolated sandboxes, maybe even lightweight dedicated diskless VMs.

[druck]

Re: Now end Windows

It's too late, just like when I left a brace of pheasants in the boot of my car for a week in high summer. Even after being stripped down and professionally cleaned, you could never open the boot again without remembering the stench, so I had to get rid of it and move on to something else.

[CapeCarl]

Re: Now end Windows

You had to re-boot?

[Anonymous Coward]

Re: Now end Windows

Re-car, sounded like; but yours is much funnier.

[Bartholomew]

Re: Now end Windows

No mention of telemetry ?

Well obviously the telemetry. I mean, the telemetry goes without saying, doesn't it?

All right, but apart from the GUI, the networking, WSL, .NET, unwanted background processes (wasting power), unwanted install time applications, telemetry and their current CEO, what else should Microisoft remove (or at least make optional) for us?

[Adair]

Re: Now end Windows

'Windows From Scratch' - it could be a thing.

[DJV]

Re: 'Windows From Scratch'

Well, that should have been ReactOS - but the chasm between what that supports versus the current version of Windows unfortunately seems to get larger every day.

[joeldillon]

Re: Now end Windows

WSL1 literally was an NT kernel personality, as was intended for NT to support from the start - and it wasn't that great, for compatibility and speed reasons, which is why WSL2 came along with the whole hypervisor thing.

Personally especially now there's X11 integration I find Windows 11 with WSL2 to be a pretty usable Linux, it's my usual development setup.

[Liam Proven]

Re: Now end Windows

[Author here]

> WSL1 literally was an NT kernel personality

That is what I thought myself, and I said so in an article. A lot of Microsofties got very angry with me on Twitter, published various insults and threats and no citations or evidence at all. To their shock and bafflement I rejected their arguments from authority and refused to back down until they refuted me: meaning, provided evidence that I was wrong.

Which, to be fair, they reluctantly did, linking to various internal MS documents, presentations and things which I had been unable to find from web searches.

To cut a long story short: no, it wasn't. It was unrelated to the NT POSIX personality except inasmuch as it came from the same company, a quarter of a century later.

A fairer summary of WSL1, as far as I can place the history, is this:

1. Some MS exec thought that having Windows Phone able to run Android apps would be useful.

2. Windows Phone being based on NT by then, MS had to develop an Android emulator for Windows NT.

3. This proved harder than anticipated, and a special version of Android's originally-a-sort-of-JVM was not enough.

(Aside: Android 1 & 2 had a sort of JVM. Android 2.2 replaced this with Dalvik, which did JIT compilation. Android 4.4 added a new runtime, ART, which does AOT instead of JIT, as an alternative to Dalvik. Android 5, ART replaces Dalvik.)

So MS had to implement an emulator for the Linux kernel API to support the ART runtime.

4. Since now, after a *lot* of work, MS had a Linux kernel emulator for NT, someone decided to offer this as a new Linux compatibility feature.

5. It worked, but not that well and performance wasn't great for filesystem-intensive stuff. Git is *extremely* FS-intensive. Linux developers love Git.

6. MS introduces WSL2, which discards the emulator and runs a real Linux kernel in a VM.

7. MS discontinues the Windows Runtime for Android. https://learn.microsoft.com/en-us/windows/android/wsa/

Summary:

In my cynical suspicious view, all because some exec with no technical knowledge promised something that turned out to be very hard and became irrelevant once another exec decided to kill Windows Phone, just about the point when it became viable to take an Arm phone, add a keyboard, mouse and screen, and use it as a PC. As a result MS spent about a decade and probably $tens of millions on a futile excursion which is probably one of the biggest OS dev exercises of the century so far.

Summary of the summary:

There are few real deep techies left in authority in MS any more, and decisions are made without proper tech understanding.

Saying that, much the same is true of Apple and indeed the enterprise Linux vendors.

[joeldillon]

Re: Now end Windows

I wasn't asserting it was a development of the POSIX personality as such, which as we know was a fairly useless tickbox exercise, just that it used the same mechanisms just as e.g. the OS/2 personality did.

But you've done way more research on this than I have, so in your case I am happy to bow to authority. :)

[Bebu]

Re: Now end Windows

Interesting history of WSL1. I too assumed it was based on the NT posix compatibility layer.

I recall a something like services for Unix or somesuch which used something called Interix. I assume it was based on the NT posix layer with Unix compatiibilty libraries.

Fortunately never had a lot to do with Windows but did have Cygwin X11 for Windows users that required X11 and some Unix compatibility (mostly ssh I think.)

[molletts]

Re: Now end Windows

I have a vague recollection from probably about 20 years ago of someone demonstrating actually using the NT kernel to run Gentoo. I don't mean running Gentoo Linux in a WSL2 VM or having a Portage-managed userspace under Cygwin, but actually using it as the kernel's userspace - booting the NT kernel then presenting a CLI at the HAL console. Presumably, the old POSIX subsystem was a major enabler, providing enough of a libc to get the basic stuff running.

It was very experimental and really only one those "because we can"-type things, like the recent demonstration of installing Linux via the Windows installation environment, but it was cool nevertheless.

[Nematode]

"Gentoo Linux to drop Itanium support as Funtoo fork enters 'Hobby Mode'"

The only words I understood there were Linux and Support. The rest suggests what I've always thought, that the Linux 'project' just continues down the road to an even more unholy mess. If I had any confidence in Linux's future, I'd migrate SMWBO to Mint to get off Windows dependency, but when I'm deid and gone, the chance of her ever finding someone locally to be able to fix stuff, you know, "open a terminal window and type SUDO..." is nil.

[42656e4d203239]

>>the Linux 'project' just continues down the road to an even more unholy mess

Nah - you misunderstand. The Linux project is ruled with a rod of iron wielded by his mouthyness Linus Torvalds. Linux is just the kernel. What seems to be a tangled forest with unrestrained undergrowth is the seemingly endless collection of "distributions"... which is both a FOSS strength and a weakness.

Strength because it allows anyone to start a distibution publish it and have people contribute (or not), weakness because how does someone, unskilled in the art, make a selection?

The above strength does also mean the most adequate will survive (who knew Darwin applies to Liunx distros?) which is perhaps a weakness as well but the alternative would be one distro to rule them all and there we are back in iOS (essentially a BSD distro isn't it? can't remember)/Windows land which is undeniably less than ideal but usable for 90% of use cases (it could be argued that its Darwin again).

The above weakness is pretty bad if you want 'everyone' to use a Linux distro intead of Windows; who tells joe which is best for him? he certainly isn't going to try them all or even a few. Joe is quite likely to pick one, find it is 'too hard' and buy an iPad/Windows box instead.

With respect to SWMBO - you will find remarkably little needs sudo in Mint, once it is set up it just keeps trucking... its definitely not beyond the wit of anyone to learn! Of course, if you decide to equip SWMBO with nVidia graphics, and any Linux Distribution, they will have to learn a lot more (and they are hostage to nVidia bothering to support your card... )

[Liam Proven]

[Author here]

> If I had any confidence in Linux's future, I'd migrate SMWBO to Mint to get off Windows dependency

Nah. ChromeOS Flex, on an old Thinkpad. Job done. Maintenance: zero.

[dmesg]

I support Windows 10/11 for a client, and I support my wife's Linux Mint install. Far less work for the latter.

[keithpeter]

@Nematode

"migrate SMWBO"

Chromebook?

Or is there specialised software needed?

[ If Slackware goes South, I'll be using a Chromebook for interwebby stuff so around 80% of screen time. Add an old Thinkpad with OpenBSD for the other 20%, maths stuff and occasional graphics / audio ]

(May your life be long and joyous &c)

[chuckufarley]

I used Funtoo for a few years...

...But it became another chore because Mr. Robbins did not commit to keeping things stable. Features were introduced and a while later removed, the "Wolf Pack Philosophy" was a bad social experiment, and I had bug reports that languished for years. Literally years. The distro was fickle because the leadership was fickle. I don't miss it to be honest. I miss what it might have been.

In the end I think the only thing that made Funtoo superior to Gentoo was the use of the Debian kernel to make installing and maintenance easier. Even that had problems though because you can have kernel level support for a feature but with out up to date user space utilities it doesn't mean much. RIP Funtoo and what it what it could have been.

[johnny-mnemonic]

The absolute final nail in the coffin of Linux/ia64 - or is it? Yeah, maybe not...

@Liam Proven

"The Linux kernel nearly removed Itanic support in February 2023, and finally did so in kernel 6.7 last October. <b>As we predicted the following month, nobody has stepped up to maintain out-of-tree support.</b>"

So nobody has stepped up to maintain out-of-tree support? Please do your research better: https://forums.theregister.com/forum/all/2024/04/12/gcc_15_sinks_itanic/#c_4847166 gives you all the information you need to get this part right.

Out-of-tree support for Linux/ia64 is there and maintained (Tomas' fork was created beginning of November 2023 already, see "created_at" in [1]), as well as for the glibc. Just check the repos at:

https://github.com/linux-ia64/

For per Linux mainline (release) candidate source code you can refer to:

https://github.com/johnny-mnemonic/linux-ia64

[1]: https://api.github.com/repos/linux-ia64/linux-ia64

And again: Linux mainline and stable (release) candidate kernels are regularly tested on actual hardware (rx4640, rx2620, rx2660, rx6600 and rx2800 i2) and in ski under x86_64. For the latter you can check for example the results on [2].

[2]: https://github.com/linux-ia64/linux-stable-rc/actions

[johnny-mnemonic]

Gentoo drops ia64 because...

"It had little choice in this. Like any other distro, Gentoo relies on upstream support for a platform in order to keep it working."

That's interesting, because there are still at least two Linux distributions I know of that support ia64 with out-of-tree patches: T2/SDE ([1]) and EPIC Slack ([2]).

[1]: https://t2sde.org/

[2]: http://epic-slack.org/

Of course one can argue that technically we that maintain the out-of-tree support for Linux/ia64 since late last year are upstream now. But even then, why does Gentoo not use our patches if we are technically upstream?

Well, I think the main reasons here were (1) that their development box (a rx3600 IIRC) "died" (see [3]) or became too unstable for continued use (see [4]) and (2) that nobody there was really interested in the architecture (see [4], [5] and [6]). It's really sad that nobody was able to fix their machine or maybe even tried - or even knew about these problems. Debian dropping support for ia64 should have freed their build and development box(es). So hardware could have been available, but maybe just not at the right place or continent.

[3]: https://public-inbox.gentoo.org/gentoo-dev/[e-mail]/T/#m07b76ff4bc5fbe352563d7bc43eaf05acd530eab

[4]: https://public-inbox.gentoo.org/gentoo-dev/46472341.fMDQidcC6G@pinacolada/

[5]: https://public-inbox.gentoo.org/gentoo-dev/[e-mail]/

[6]: https://public-inbox.gentoo.org/gentoo-dev/[e-mail]/