A Proposal To Go 64-bit Only With Fedora 23

[gens]

fedora and redhat guys are like this always, no big news here, you can always use something like like centos 6 or debian 8 or ubuntu 16.04 with lxde or xfce desktop for many years don t be mad because some stupid guys

fedora wants over 700MB ram to install (xfce), so their goal is clear

redhat (as in redhat) people are usually very competent

[gens]

The P4 was hot. The Willamette P4 was on the toasty side of hot, but it was still reasonably manageable I don't know if those would be very useful today, especially ones running Intel IGP graphics of that time period. The Prescott P4 was on the Fire and Brimstone side of hot, I doubt many of those motherboards survived this long.

yes, older ones like willamette with its 180 nm lithography
"newer" ones using 130 nm down to 64 nm don't get that hot, especially the celerons

also granted that was before todays power saving things
on the other hand they are single core, that might lower the relative (lithography wise) power usage a bit compared to todays cpus

[Anvil]

RHel7 has dropped some if not all 32bit Arch, so i think this would be a good idea

[eidolon]

...no official proposal has yet to be submitted and likely wouldn't be approved by FESCo for Fedora 23.

It seems a little early for people to froth at the mouth. But by all means, people should voice concerns if they have them via the appropriate channels.

A few comments from the blog post:

• ...a larger number of the i386 [yum usage] are older releases no longer supported by Fedora
• ...my proposal would not be about ending making x86_32 or arm_32, but moving them back to secondary architectures where they would need to have teams of people interested in them doing the work.
• ...it is very hard to count ARM users to know how many people it affects. From yum daily checkins it would be less than 400 computers versus 10,000 x86_32 computers. I realize there are all sorts of reasons that metric doesn't cover how many real ARM systems might be there, but it is the only one that is around. If I am going to make a horrible proposal that inconveniences 10k x86_32 users, 400 ARM boxes would also fit the list of being moved out of primary.
• I seem to have over-emphasized the 'dropping' of 32 bit in my proposal. I am not saying Fedora would stop making them, but that they would be a secondary architecture which means that they would need to rely on a separate build system and would need a team of people who are interested in the architecture to run that group.
• I would expect that 32 bit cross library builds of packages would still occur on the x86_32/x86_64 architectures, but that the work on getting the installer and a self-hosting OS from those packages would not be done unless a secondary architecture team did that work.

Among the comments on the blog, a developer from Sugar Labs commented on XO deployments, for which Smoogen didn't have a great reply. The XO-4 and XO-1.75 are ARM-based (ARMv7), while the XO-1.5 and XO-1 are x86-based.

[eidolon]

RHel7 has dropped some if not all 32bit Arch

For what it's worth, Springdale Linux 7.0 (RHEL 7-derivative) supports x86 architecture.

[Aguazul]

I think abondoning 32-bit ARM seems very unwise, but 32-bit x86 should die really.

I run a 64-bit Debian system, but I run 32-bit executables for Firefox and Google Chrome using multi-arch. The memory use of browsers really is almost half when running 32-bit. This makes a big difference with a lot of tabs open. I think there is an excellent case for running 32-bit executables for certain apps when you don't need to access multi-gigabytes of RAM in any single process.

[Styromaniac]

I'm all for dropping 32-bit ARM if it suits them. New ARM devices are shipping with 64-bit CPUs if my crappy $100 phone is any indication. (It's only crappy because the 4GB internal storage cannot fit all this carrier bloatware)

For x86 though... I would like to have the option to install Fedora 23+ on 32-bit machines I guess. I can't see myself doing it because I only have a 64-bit x86 CPU, but if someone else can breathe more life into an aging PC and it happens to be Fedora, then there's more Linux adoption potential.

By the time Fedora 23 is out though, you could likely buy an SoC with either x86-64 or ARMv8 for pocket change and be happy with it.

I don't know the best solution, but hopefully the least painful/costly route is taken between developers and users, whatever that may be.

[eidolon]

...it is very hard to count ARM users to know how many people it affects. From yum daily checkins it would be less than 400 computers...

To add, Smoogen's Yum data does not reflect the One Laptop per Child XO-4 & XO-1.75 (armv7hl) users or Pidora (armv6hl) users out there (however many in number they may be), but I'm not sure the extent to which the Fedora Project owes those projects accommodation if their developers aren't actively involved in the Fedora Project itself. Hopefully a path forward can be reached that everyone is content with.

[zxy_thf]

Don't kill x86_32!

1. There are still a lot of good 32 bit x86 machines around.

It's really difficult to find good old (REALLY OLD) i686 computers, which means old Pentium or Athlon(without -64 suffix) computers.
I don't think such a system will have sufficient memory to run modern desktop distro smoothly.

I would definitively use CentOS 5 rather than Fedora on such systems.

[lovenemesis]

Don't kill x86_32!

1. There are still a lot of good 32 bit x86 machines around.
2. There is a plethora of 32 bit software around.
Actually I had a recompile of a system these days to support Gentoo multilib. It was neccessary due to a mesa version in the precompiled 32 bit compatibility libs that was too old.
But I noticed how much software (especially binary stuff) makes use of these 32bit libs. So if you'd kill off 32bit now tons of software would stop to run. And people are arguing that "Linux isn't a platform" cause you can't get commercial software to run on it. This move would not improve the situation.

It won't kill your 32bit software as long as multilib continues to be alive.
RPM distros like Fedora handles multilib quite well so far.