Suggestions please - Office for Linux

Suggestions please - Office for Linux

Post by Rod Smi » Wed, 20 Aug 1997 04:00:00



[Note that I've set followups to comp.os.linux.hardware.]





>: >If we don't get support of new hardware, it will be more and more
>: >difficult to find a Linux-compatible computer.  One of the reasons I
>: >used Linux in the first place is because it had support for all of my
>: >hardware.

>: I agree. What we need is support for plug and play. You will probably
>: say Plug and Play sucks, but it works fine on my win 95 installation.

> See, this is exactly what I"m talking about.  WHY do we need plug and pray
> support?

For the reason implied by Mark Lehrer, quoted above -- because, whether
you or I like it or not, new non-PnP hardware is becoming a rarity, so if
Linux DOESN'T support it, it'll become harder and harder to buy or build a
Linux box.

It might also be useful to distinguish between two different types of PnP
hardware: ISA-bus and PCI-bus.  (I don't even know if the term "PnP" can
technically be applied to PCI boards, but the design goals and mentality
are similar, so I do.)  Since PCI was designed with this sort of thing in
mind, PnP-type features are easy to implement on it.  ISA bus was never
designed for it, and much of the controversy about PnP centers around ISA
PnP components.  I say it's useful to split the discussion between these
simply because ISA bus is on the way out.  I'm sure it'll still be
present for another year or two, at least, but there are increasing
(though still fairly small) numbers of PCI-based sound cards, modems,
etc.  That trend will only accelerate with time, especially if there's
truth to the rumor I heard that the "Designed for WinXx" label can now
only be applied to machines with no installed ISA components.  Anyhow,
the whole ISA PnP thing will become a bad memory before too terribly
long, leaving us with PCI and anything new the industry decides to throw
at us.

Quote:> It just increases the size of the kernel.

Only if you compile in support for it.  I admit I haven't checked the
latest development kernels, but I find it hard to believe that it's an
absolutely necessary part of the kernel.  Of course, if you've got PnP
devices, leaving the PnP support out may not be an option; but if you're
so opposed to it that you'd prefer not to see that support at all, you've
presumably not bought any PnP devices, and so don't need it.

Quote:> It's helping along
> those companies that are manufacturing "non-standard" or "incompatible"
> hardware.

AFAIK, there's little or no correlation between PnP and the most
notorious of the non-standards out there -- the so-called WinModems and
GDI printers.  In fact, since printers aren't ISA-bus components
themselves, they aren't really affected by PnP support, AFAIK, or at
least not support for PnP ISA cards.  There are plenty of PnP modems that
are NOT WinModems, and I can think of some bizarre non-standard things
that predate PnP (like modem/sound cards based on IBM's Mwave DSP).

Quote:> Basically, you're helping MS if you buy plug and pray crap.

Does Microsoft get royalties on PnP hardware?  I know they were involved
in the specification, but I don't know if they profit directly from it.

Quote:> Besides, plug and pray is not known for its quality...

From what I gather, PnP itself works fairly well FOR CONVENTIONAL
CONFIGURATIONS.  Joe Schmoe adding a PnP modem to a PnP-aware
pre-assembled system built by a "name" manufacturer after PnP became
common usually doesn't have problems.  It's when you get into unusual
configurations, like PCs with SCSI adapters or scientific data-acquisition
boards or whatnot, that you get problems.  Especially when you throw in
the Win95 tendency to insist on configuring the hardware the way IT wants
to, without allowing overrides.  My understanding is that the existing
Linux PnP support gives the user full control, so this last bit isn't an
issue for Linux PnP.

If you're claiming that PnP hardware tends to be of lower quality than
non-PnP hardware, well, that's a statistical correlation and will have
plenty of exceptions on both ends, so I don't think it's a useful basis
for deciding what Linux can or "should" support, even if the assertion is
true, which you haven't demonstrated.

--
Rod Smith                                 Author of:

http://php.indiana.edu/~rodsmith          "OS/2 Soundcard Summary"
NOTE: Remove "uceprotect" from address to e-mail me

 
 
 

Suggestions please - Office for Linux

Post by Tony Langd » Fri, 22 Aug 1997 04:00:00


It's 20 Aug 97  06:22,

discussion of Suggestions please - Office for Linux

 ro> It might also be useful to distinguish between two different types of
 ro> PnP hardware: ISA-bus and PCI-bus.  (I don't even know if the term
 ro> "PnP" can technically be applied to PCI boards, but the design goals
 ro> and mentality are similar, so I do.)  Since PCI was designed with this
 ro> sort of thing in mind, PnP-type features are easy to implement on it.

PCI usually gives little trouble.  Actually, I had my first PCI PnP
related problem today (Under Windows 95 BTW).  The BIOS configured a
SCSI card and a network card on the same IRQ.  Seems they didn't like it
(thought PCI was supposed to be able to handle IRQ sharing).  I fiddled
around, and the next time, the network card was shared with something
else.  This time, everything worked. :-)

 ro> ISA bus was never designed for it, and much of the controversy about
 ro> PnP centers around ISA PnP components.  I say it's useful to split the

I agree.  I don't like ISA PnP myself, and disable it where possible.
It's great on buses designed for it though.

 ro> Only if you compile in support for it.  I admit I haven't checked the
 ro> latest development kernels, but I find it hard to believe that it's an

That's the beauty of Linux.  You can pick and choose what support you
want, and whether to make it part of the main kernel, or as a loadable
module (great for occasionally used devices).  One of the first things I
do after installing is to recompile my kernels. :-)

 ro> absolutely necessary part of the kernel.  Of course, if you've got PnP
 ro> devices, leaving the PnP support out may not be an option; but if

True, but there's possibly a lot of other things you can leave out
instead. :-)

 ro> From what I gather, PnP itself works fairly well FOR CONVENTIONAL
 ro> CONFIGURATIONS.  Joe Schmoe adding a PnP modem to a PnP-aware

Not always. :-(  I had a case where a PnP ISA sound card and a PCI
device had an IRQ conflict, which crippled the system.  The BIOS
couldn't assign the two different resources or disable one of the
devices.  I think I tried some fiddling with the BIOS, and eventually
just pulled out the soundcard to avoid any hassles.

The only trouble free setup I've seen is all PCI, which usually works
well (except for the one off SCSI conflict I mentioned above).

 ro> hardware the way IT wants to, without allowing overrides.  My
 ro> understanding is that the existing Linux PnP support gives the user
 ro> full control, so this last bit isn't an issue for Linux PnP.

This helps, as does a BIOS which allows you to reserve some IRQs for
legacy cards, and leave the rest to PnP. :)

... Biology grows on you.
--
| Fidonet:  Tony Langdon 3:632/367.2

|
| Freeway Internet Gateway, Melbourne, Australia.  For information about
| our services and conditions of use, e-mail: