Re: Architecture baseline for Forky

On 2025-10-28 09:39 [email protected] wrote:
> 
> yes - maintaining all special cases is an effort.
> 
> ... 
> 
> yes, but your mileage may vary: You often have a special case where general
> performance measuerement didb't catch the use case.
> 
> ... 
> 
> yes - that are all people involde in debian or generally in Open Source.
> 
> ... 
> 
> yes, throwing out old clumsy code with special workaround for old hardware flaws
> is a great cleanup.


First of all: I know that it is not my place to comment on that. I am 
not a developer, I only use Linux system at home for some ~30 years now. 
But it might be interesting to read my perspective. Obviously I cannot 
speak for production environments and business use cases.


Please feel free to simply ignore this message if you do agree that it's 
NOT my place to comment here... Thank you!


If you do read on:

My primary question is this: for which group of users is Debian supposed 
to be in the first place? If it is for the users who already have 
certain hardware, then keeping existing hardware supported should be one 
of the goals of this Linux distribution. No?

Secondly: off course it would be cleaner to through out older code that 
supports older hardware. But this is not the point, considering the 
primary question: for whom do you make Debian?


For me it comes down to the amount of effort required to keep existing 
hardware running. If it is very tedious and time-consuming, there will 
come a point in time when it is no longer feasible. Take x86-64v1 for 
example. Keeping it running should be relatively easy. Take other 
systems like Big Endian PowerPC for example, and the amount of work that 
is required to keep it a supported architecture is much much more.

So, the questions should be: 1) how easy or hard is it, to keep the 
architecture supported? 2) how many systems are still in use that 
require support? In that order.

(And 3) is there a Debian maintainer who still want to put effort into 
keeping it around.)

I come from Gentoo Linux, and when people say that it's an issue on 
which x86-64 version a Linux distribution should be based on, I raise an 
eyebrow. Naturally, because on Gentoo Linux this is NOT at all a 
question. But for all other Linux distributions, off course this is 
important. Base it on v2 and you lock out so many still very usable v1 
computers. But if there is so much to gain if you were to switch for 
example to x86-64v2 or even v3 or v4, this IS something to consider.

If it is easy to provide multiple builds, wouldn't it be also possible 
to provide a very low baseline package for x86-64v1, and then for only 
selected packages that really benefit from it to additionally provide 
v2-v4? For software that doesn't gain in performance at all you wouldn't 
have to bother...

For POWER, specifically for my Power Mac G5, Debian is my last hope. I 
did try Gentoo on this machine, but for me as a user (yes, maybe an 
advanced user, but not a developer) I did struggle a lot on those older 
machines, because software itself is loosing compatibility with this 
architecture. (Released versions are mostly no longer tested for 
POWER4/5 and Big Endian, so very often patches are necessary, which take 
time, and more than occasionally things break, bringing the necessity to 
report a bug, and this would leave me constantly reporting build-issues 
on this specific hardware... Or it leaves me with legacy/older versions 
that still function. On Gentoo, a rolling release, this is tedious. 
Which would not the case if I used Debian/PPC32/64 as a user, because 
you developers and maintainers then fulfill this task.) This is not the 
case with x86 mostly, so even x86-32 should still work better than PPC32 
or PPC64. Debian, I hear, does a great job keeping PPC32 and 64 alive, 
but I can only imagine how much work that is.


Still, the hardware exists and will continue to function until it breaks 
or I (and everybody else), though still functioning, throws it in the 
trash. This has something to do with planned obsolescence, but at this 
age (>20 years for the G5s) not so much as one might be inclined to believe.

So, Debian is NOT for enthusiasts. But if some older architecture is 
easy to maintain, why drop support (e.g. the x86 case)? Even for 
enthusiasts (like early x86-32 hardware), though not the main audience...

As for POWER8, as I understand it, there are still so many users around. 
This shouldn't be a discussion about clean code then... If you wanted 
that, you'd be better off with Apple. They constantly "clean" their code 
and remove support for older hardware, for no apparent other reason than 
their own business model. Or you are a cutting edge Linux distribution 
that specifically targets the latest and greatest. Arch Linux for 
example is only available on x86-64, but even they base their binaries 
still on the 20 year old baseline "v1".[1] BTW, they did /some/ 
benchmarking back in 2020 whether x86-64v3 brings some performance boost 
or not.[2]


These are just my 2¢. PLEASE feel free to simply ignore all of this.


Greeting,
Linux User #330250 for more than 30 years now.
Please don't crucify me for not being 100% a Debian user...


[1] https://bbs.archlinux.org/viewtopic.php?id=302425
[2] https://rfc.archlinux.page/0002-x86-64-v3-microarchitecture/