Quoting Steve Litt (sl...@troubleshooters.com):

> There are a couple kinds of specialty hardware:
> 1) Antiquated stuff
> 2) Newish stuff that just might make it

Granted, and a fair point.  So, I'll clarify that for my own present
discussion, I've been talking entirely, exclusively about category #1,
antiquated stuff.  E.g., the aforementions PowerPC variants including as
a minor and somewhat comical example my G3 iBook.

(I belatedly looked down, Steve, and saw that we do concur about
category #1.  So, the following is _not_ aimed at you, and thank you for
the additional supporting sanity.  Please indulge me while I drive my
point rapidly into the ground.  I hope the following is amusing reading.)

If one wishes to talk separately about category #2, anyone who so wishes
is welcome to proceed on that subject separately without me, as I am
talking about antiquated.

Exemplars of antiquated designs, being both physically very old and
(typically) no longer in stock hardly anywhere, are fragile and can be
expected to have very limited and unpredictable remaining service life.
As they (inevitably) develop various problems inherent in old computer
hardware, the owner finds to his/her displeasure that replacement parts
have now become a _specialty_ market, hence more expensive than
previously in constant-per-year value, not to mention the equipment now
being (IMO) woefully deficient in quite a variety of ways compared to
more-recent and more-standard kit.

There was a day when the writing was on the wall that nobody with any
common sense would -- with probable rare and exotic exceptions -- put
any more money into PPC-anything-based systems, and that was around the
time that not just Apple but pretty much everyone else gave up on it
and, with a heavy heart, migrated to x86_64.  If memory serves, that was
right around the year 2006, when even diehard G5 PPC fans gave up.

My G3 iBook dated from circa 2002 CE.  GNU cal tells me that this lovely
present day is in the year 2018 CE, sixteen years later.  If this has
been an automobile, it would analogously be a Datsun (the name of Nissan
in the USA from 1956 to 1986, when the USA joined the rest of the world
for a change).  It would be a car with sparkplugs and bias-ply tyres,
because radials were too new-fangled.

In 2018, the rational attitude to PowerPC, IMO, is indulgent laughter
and 'Then we let it go.'  And it would IMO be the height of folly to add
_any_ of the we-now-know innumerable, mutually incompatible and
gloriously irrelevant PPC variants as a supported architecture to a 2018
Linux distribution.

I'm old.  But I'm not that stupid, and I hope Devuan Project isn't,

OK, I lied.  I respect your views, Steve, so I'll say a couple of words
about category #2.

> Examples of #2 are
> various Raspberry Pi's, Beaglebone, various modern SOC's.

This is certainly _less_ outright unreasonable, but as a generality,
each of these is to a large degree a special snowflake.  Special
snowflakes are time sinks.  IMO, a projecct in Devuan's present age and
condition should be wary of those, as each is very, very likely to
divide resources and complicate planning for the project.

You probably remember what happened to Debian Project when the supported
architectures reached a totally absurd quantity, right?  Something like
14 at the maximum?  The project as a whole started becoming even more
unwieldy than it had been, and frequency of releases became so rare that
it became a standing joke all across the Internet.  Debian was, among
the technical problems, caught in (one of many) political snarls that
then took years to unsnarl.  This should be a cautionary tale:  Add
special-snowflake platforms, no matter how trendy, with extreme caution.

And yes, RPi is a special snowflake.  kernel.org kernels _still_ cannot
run on them, large and specialised and obscure out-of-tree patchsets are
requires, a special bootloader is required, etc. -- qualities RPi so far
shares with every other ARM variant of any description.

Some day, inshallah, that may change.  That day is reportedly not today,
and tomorrow seems unlikely, too.

'Various modern SOEs?'  Please cite even one of those that runs any
current kernel.org kernel, and works with anything you can call a
reasonably standard bootloader without breaking out laughing.  OR, hey,
can you name one of which even _one_ of those two things is the case?
I'll believe it when I see it, and I'm not going to hold my breath
waiting for an example.

Do specialty products whenever the project seriously and soberly doesn't
need the time and resources for something more important.  I think you
will find that that time will be a long distance from now.

Dng mailing list

Reply via email to