For me personally only really the x86_64 architecture matters, I do have an ix86 machine sitting under my desktop but it is humming along nicely. Everyone really supports these just fine, but here is a thing I don’t get. Why do so relatively few distros support ARM and MIPS. Firstly let’s examine why these archs are important, ARM is widely predicted to take a huge bite out of the netbook market, a segment it is extremely well suited for. MIPS naturally is important because it is the foundation for the Chinese YeeLoong chips which we are likely to see the Chinese market gravitate towards. Being a market of a billion potential users with additonal opening to the west via netbooks this is definitely a place Linux distros should be putting some effort, we have literally no competition here, the first distro to integrate really good support will likely be able to pick up millions of new users over a short span of years. This is also a good target because this specific segment has shown signs of being the first truly open platform we might get, from the BIOS and up.
Having a significant chunk of the netbook market is likely going to be important. It is a good market for Linux as the use cases for a properly promoted netbook do not call for “regular” desktop tasks but more likely a specially designed interface such as is see with the EeePC or the Aspire One.
Thus in my search for a new distro to replace Fedora, one requirement I will set up will be support or planned support for ARM and MIPS. Not because I personally need them right now but because it is my belief that an important piece of the future for Linux lies in this direction.
http://www2.apebox.org/wordpress/wp-content/gallery/00-single/armless.png
Yes, indeed!
Debian includes MIPS (both big and little endian) and little endian ARM amongst its core arches.
Ubuntu builds for ARMel, but not MIPS (though officially only supports i386/amd64)
openSUSE has pretty much abandoned support for “odd” arches.
Fedora states that they want to support ARM, but don’t offer little things like a kernel.
Arch is i386/amd64 only.
Gentoo appears to do both MIPS and ARM, but details are missing from packages.gentoo.org
Foresight appears to do only the arches that Arch does.
At this point I get bored & give up searching
When I say “odd” I mean “no MIPS or ARM, but we do ia64″, in case that was unclear
I never quite understood the point of putting so much effort into SPARC, ia64 and PPC/PPC64. I am not thinking there is a whole lot of that hardware in community hands also the future does not look all to bright for those platforms. Certainly Smolt doesn’t really show impressive stats for them. For my money time would be better invested in ARM and MIPS, but then again we should really do the Debian thing and support everything the end result is better software even if it is a lot of work. Also we can’t tell volunteers what to work on, with something like Fedora’ arch teams unwitting packagers such as myself can even get help sorting out problems on these less common archs.
I knew that Debian has excellent arch support, NetBSD might run on a toaster but Debian runs on everything relevant (even irrelevant things like HURD as well). It’s truly very impressive and should be applauded.
There was a conference[1] not long ago when ARM presented some very sexy netbook and nettop hardware which they specifically demoed with a Ubuntu 9.04 snapshot on. I have to admit after seeing that stuff in action it was hard not to be impressed.
I used Gentoo for years, it’s very slick but compiling and finding just the right switches for every package on the system is a bit beyond me these days. I like a system like Fedora where everything is neatly packaged. Never the less that is an option.
I know that at the last FESco elections there was a big push for secondary arch support but nothing really seems to have come of lofty goals for any of the candidates. I would love to see Koji start building for all the secondary archs we can muster build machines for, just to see how much work it would be to fix up the distro and pipe patches back upstream.
openSUSE has SoC goals to support ARM but also recently announced no plans of their own to support that platform. They do support PPC/PPC64 so hopefully we will see that coming in the near future.
Outside of that I think we are some what out of luck, it is a lot of work to support and requires access to hardware. Outside of the main distros I don’t think those resources really exist.
[1] http://news.zdnet.co.uk/hardware/0,1000000091,39617753,00.htm
PPC/PPC64 is quite important to run Linux virtualized on big IBM boxes; SPARC design is open-source (see OpenSPARC).
Fedora and OpenSUSE probably has the same dilemma in terms of ARM support: their parent companies, unlike Ubuntu/Canonical, are not heavily invested on personal desktops/laptops, and as such, not much resource is pumped into it. Fedora’s ARM support is mostly the work of embedded developers, AFAIK — this might change once the first ARM netbook with a decent amount of RAM comes out.
I think the only proper way to do this would be to have build machines for all of them. If something happens to fail on anything !x86_64/ix86 then it could be considered non-fatal. Most package maintainers really don’t know then intimate details of say the ppc platform.
Take Mono e.g. 2.4 fails on ppc, Paul has no proper idea of why, when asking for help he is told that Mono basically does things wrong but nobody probably investigated the issue in depth and produced a patch. This is where we need architecture teams to lean on.
With my Fedora maintainer glasses on (even if I retired them), I really want to support ppc but I don’t have access to the hardware nor do I have the required knowledge. I can however go out and get an ARM or a MIPS based netbook (actually I have ARM in my pocket, thrice my nokia n810, my phone and my calculator), I also predict that the amount of this hardware in consumer hands will increase.
This is important because we have no competitors there, we can for the investment in a bit of hardware and appealing to interested people gain a significant advantage. One that will benefit everyone since bugs fixed for portability reasons will probably be bugs on all platforms as it is. It would also give us an important way to give back to upstream, we have access to an open transparent build system and the hardware. Something that can’t be expected of every project out there, just telling developers where their code compiles and not would probably be useful.
Aside that just like with Mono’ position in Fedora, I worry whenever Red Hat’s business incentive affects where Fedora could benefit from going. PPC is it is meant mostlly for big iron IBM shininess is not a good primary target for Fedora. With OLPC investigating ARM as well as netbook vendors everywhere this makes a much better default as our 3rd arch to be supported. I don’t think our sponsors business model is a very compelling reason for determining policy for Fedora. Not that I don’t wish Red Hat all the luck in the world and celebrate their continued success but Fedora is a community project and as such I think we should make decision to benefit Fedora first and RHEL second.
That being said, I think we should support everything, Fedora is a vibrant community and I am sure if we can manage to get the hardware merged in our buildsystem then it would be valuable to everyone to have this going. I would happily pitch in money to buy the required hardware, I believe we already have people interested in making this happen with knowledge to help work out bugs. I don’t think it would be fair to sit back and expect Red Hat to pay out for everything, even if they benefit from the end result as well.