On Tuesday 08 of September 2015 13:13:12 gumb wrote:
So in March 2013, the proportion of 32-bit downloads was 40%. Whilst overall downloads almost doubled with the 13.2 release, the *proportion* of 32-bit downloads halved. If that had read 'number' or 'total', we could assume that 32-bit had not just halved in number but now represented almost half of that again due to the overall user base doubling, but by saying 'proportion', that implies the percentage of 32-bit for 13.2 was around 20%. Sure, let's give or take a bit and accept that we're looking at a new release in two months from now, one year after 13.2 came out. So we might be looking at 10 to 15% using 32-bit machines come the time of the Leap 42.1 release.
The difference between downloads and users aside, you are still mixing two very different numbers: number of users with a 32-bit system and number of users with a 32-bit hardware. Even some of the most passionate advocates of keeping an i586 release admitted that they are actually using i586 openSUSE on a 64-bit capable hardware for various reasons.
It's nonetheless surprising that this architecture is being dropped so soon. Take a look at the options in /usr/src/packages/RPMS/. (open)SUSE is now chasing the 64-bit ARM market, has support for ppc, sparc and more. All of these which add up to perhaps a couple of percent of the market. Sure, 64-bit ARM is growing and in two years from now, nearing the end of the Leap 42 lifecycle, it might be considerably more relevant. But this is an odd offering overall. openSUSE is effectively saying 'you can run the world's most popular architecture, or take your pick from among a bunch of obscure ones hardly anybody uses, but you want the world's number two most used architecture?
This is plain wrong. None of the architectures you mentioned (ppc64, ARM64, sparc etc.) is really official and supported in the sense x86_64 is and i586 has been until now. What is (most likely) going to happen is moving i586 to the "ports", i.e. giving it the same status as ARM64, ppc64 etc. If someone is going to dedicate their effort to making it work, that is. (As people do for the others.)
I'm grateful to Takashi Iwai for some of his thoughtful comments, including his considerations of the likelihood of future kernel support. Look at i386 support which was only finally dropped by the kernel quite recently. That was a technology around twenty-five years old, but even 486 and early 586 users had almost disappeared off the radar by the time it was finally abandoned. On the contrary, we know that 32-bit machines are still everywhere, whether that be in slow-moving corporate environments or hobbyist gadgets.
This is a very different thing. Once an architecture is dropped by upstream, there is little chance any distribution would build it any more. Dropping an architecture from mainline is therefore much more destructive step than some distributions dropping it and it only makes sense it's going to happen much later. Please note that (1) most distributions stopped building 386-compatible packages years before 386 support was dropped by mainline kernel and (2) mainline kernel IIRC supports ~20 architectures, most of which (almost) no distributions actively support (and never did).
But there are also a number of very defensive and at times rather unpleasant folk of whom I'm not sure of their motivations other than sticking the knife in to those less fortunate than themselves; a sort of baying group of bullies.
A matter of perception. Someone could rather see bullies in those who insist others are obliged to work on providing i586 support while they are not willing to help with the work (not all, fortunately, some are willing to help). Michal Kubeček -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org