older computers

Eric Battersby gyre-Ja3L+HSX0kI at public.gmane.org
Tue Sep 15 22:24:10 UTC 2009


On Mon, 14 Sep 2009, D. Hugh Redelmeier wrote:

> | From: Eric Battersby <gyre-Ja3L+HSX0kI at public.gmane.org>
>
> | Can someone clarify for me: what is an "older computer"?
>
> Probably not.  It is a generalization.  The actual attributes that
> matter are not determined by a clock.

That's the thing: they are not determined by age, but perhaps
by "era" because older parts are not the same.

> | Which new distributions should not be installed on which hardware?
> | No need to mention extremes, but what are the cut-off specification
> | ranges in terms of non-obvious attributes?
>
> What non-obvious attributes do you think matter?
>
> The x86 architecture underwent modest changes at each step from 386 to
> 486 to Pentium to Pentium Pro to Pentium II to Pentium 3 to Pentium 4.
> Some distros (and I had thought F11 was one) may require PII
> instructions (aka 686).

PII is one of the explicit requirements, so that is OK.
Non-obvious attributes include those from video cards, monitors,
PCMCIA cards, PCI cards, interactions with external peripherals.
I don't know what else, because they are not obvious.

> I have a machine that boots Fedora Core 2 (I think) and in the process
> the kernel whines that the BIOS is old enough that it doesn't trust
> the ACPI.
>
> Old machines have ISA and all that means (no sharing of IRQs,
> autodiscovery of devices is hard).

ISA might be an example of a non-obvious attribute.
I don't remember seeing bus type (ISA, VLB, PCI)
mentioned in the list of hardware requirements.


> | That threshold goes back to computers before 2000, but there
> | is *no* mention of *age*.
> | I don't think adequate testing has been done on
> | "older computers", especially for Fedora.
>
> Why would age itself matter?

It wouldn't, hence the quotes.
Hooking up older computers with newer hardware (eg: USB 2.0 cards,
DVD writers), that were not available when the older computers
were produced seems to be part of the problem.
However, the real problem is *inadequate testing*
with these new "bleeding edge" Linux distros.
I could literally find 10 bugs in F11 within one hour.
Some of those bugs are not even machine dependent.
How is that acceptable?

I have two normal standard machines,
a Dell C640 Laptop (P4, 2GHz, 2GiB), and an older Dell Optiplex GX1.
I installed F11 on both of them.
Sound did not work on either; I got hundreds of 'pulseaudio'
messages in '/var/log/messages'.  When I did get sound working
on the Dell C640 (by using 'alsamixer -c0'), it took up about
20% of CPU and died within a few minutes many more log messages.
How difficult is it for developers to do an install and
try the most basic of features?
(Since then, I have removed pulseaudio, and sound works
great.)

It would be good to have everything tested up front, right
after installation, so the user will know if he can use this
distro, rather than find out 2 months later that something is
fundamentally flawed.

Is there such a suite of regression tests that Linux can run
that will test "everything" (like video tests, sound tests, USB
tests), etc?

The latest disaster is the I cannot create a DVD in F11
with an external Plextor DVDR PX-716A writer.
It hangs after writing about 7 MBs.
I installed the same model writer in the Dell GX1 directly
on the EIDE cable.  A similar problem happens.
It hangs after writing about 7 MBs, plus a kernel process
'scsi_eh_1' is running at 23% CPU and I cannot kill it.
The external Plextor drive worked fine in F7 on the Dell C640.

I'd like to switch distros, but now F11 has me locked in :-(.

--
Eric B.
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists





More information about the Legacy mailing list