[gentoo-dev] Drivers/Portage

[gentoo-dev] Drivers/Portage

[Martin Owens]

This is probably not a new idea, but I've had the idea of creating a driver 
portage system that contains information firstly, detection secondly and 
configuration scripts and tools thirdly.

the idea is to enable admins and users to quickly learn if their device is 
suported, find out what device they actualy have and maybe get the right 
configuration, kernel modules and scripts installed.

I just know someone has to be working on this, because it seems like such a 
worthwhile project to enable Linux to escape the hardware headache monica it 
has too readly acuired.



--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Drivers/Portage

[Greg KH]

On Fri, Jul 16, 2004 at 06:57:13PM +0100, Martin Owens wrote:
> I just know someone has to be working on this, because it seems like such a 
> worthwhile project to enable Linux to escape the hardware headache monica it 
> has too readly acuired.

What hardware headache?  Linux supports more devices, "out of the box"
than any other operating system today.  And that was done without specs
:)

thanks,

greg k-h

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Drivers/Portage

[Jason Rhinelander]

Martin Owens wrote:
> This is probably not a new idea, but I've had the idea of creating a driver 
> portage system that contains information firstly, detection secondly and 

In other words, what lspci/pciutils does?

> configuration scripts and tools thirdly.

This is (sort of) what Redhat's kudzu is meant to do - detect new
hardware and automatically configure it for you, more or less like what
MS Windows does.

> the idea is to enable admins and users to quickly learn if their device is 
> suported, find out what device they actualy have and maybe get the right 
> configuration, kernel modules and scripts installed.

Again, basically what Kudzu does.  To be honest, way back when I used
Redhat and/or Mandrake, I found the automatic hardware detection
terribly annoying.

> I just know someone has to be working on this, because it seems like such a 
> worthwhile project to enable Linux to escape the hardware headache monica it 
> has too readly acuired.

I suspect Redhat's automatic detection programs annoyed more experienced
linux users (i.e. admins) than it helped; most admins are able to run
lspci in order to see what kernel configuration/modules they'll need.


If you really want to do something to help the "hardware headache" (I'm 
not sure what 'monica' means - I couldn't find it in any dictionary) - 
help the linux kernel developers to actually support more hardware more 
stably (like Serial ATA, ACPI suspend-to-ram, etc.)  The "hardware 
headache" doesn't come from having trouble identifying cards, it comes 
from having to be nearly on the bleeding edge of kernel development for 
support of any hardware less than a year old.


--
-- Jason Rhinelander
-- Gossamer Threads, Inc.


--
gentoo-dev@gentoo.org mailing list

[gentoo-dev] Re: Drivers/Portage

[Duncan]

Jason Rhinelander posted <40F82EE5.5030501@gossamer-threads.com>,
excerpted below,  on Fri, 16 Jul 2004 12:39:17 -0700:

> Again, basically what Kudzu does.  To be honest, way back when I used
> Redhat and/or Mandrake, I found the automatic hardware detection
> terribly annoying.

Fresh from Mandrake, here, and no kidding about the annoying detection. 
For one thing, if I let it do its thing, it would have killed my very
carefully constructed comments in for instance fstab.  One time I let it
do so and it munched the comments into what it thought were a whole bunch
of mountpoints, so when I rebooted, I had all sorts of new root level dirs
named things like "the", "fs", etc, all the things that had been in my
comments, since it "helpfully" created all those dirs, so the non-existent
file systems could be mounted on them. <g>

If their stupid automated tools could learn to leave comments intact, and
above the lines they were commenting about, the tools might have
actually been worth something, but..

Now.. I DID like to use their nice visual diskdrake for partitioning,
etc.. only I learned NEVER to let it write the changes to fstab, only
using it to create and format the partitions.   When it asked to write the
fstab changes (and thankfully it DID ask, gotta give them credit for
that), I'd say no, and do it manually, so I knew exactly what I had, and
it stayed in the format I wanted it.

The same thing, of course, applied to their XF86Config tools.  After the
initial generation, I never let them touch it again, because I had
painstaking manual setups in there for my multi-head xinerama thing, and
their took found that to complex to deal with.

I never tried their auto-lilo.conf editor.  However, due to the fact that
I was running an unusual setup there as well, and with the experience I
had with the other tools, that was probably a good thing.

-- 
Duncan - List replies preferred.   No HTML msgs.
"They that can give up essential liberty to obtain a little
temporary safety, deserve neither liberty nor safety." --
Benjamin Franklin



--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Drivers/Portage

[Martin Owens]

I think I ort to make myself a little clearer.

it's not about auto detection / auto configuration. it's about information 
gathering and resources, sure auto configuration is an option at a much later 
stage but I'd much prefere a system that provides information to the usrs as 
a primary goal.

The most important part is creating/generating unique id's for hardware that 
you can match the information up to. then if the system gets it wrong or 
dosn't have any information allowing the users to input new data (subject to 
review) which can then be used to expand and revise the information.

-- DoctorMO --

On Friday 16 Jul 2004 20:39, Jason Rhinelander wrote:
> Martin Owens wrote:
> > This is probably not a new idea, but I've had the idea of creating a
> > driver portage system that contains information firstly, detection
> > secondly and
>
> In other words, what lspci/pciutils does?
>
> > configuration scripts and tools thirdly.
>
> This is (sort of) what Redhat's kudzu is meant to do - detect new
> hardware and automatically configure it for you, more or less like what
> MS Windows does.
>
> > the idea is to enable admins and users to quickly learn if their device
> > is suported, find out what device they actualy have and maybe get the
> > right configuration, kernel modules and scripts installed.
>
> Again, basically what Kudzu does.  To be honest, way back when I used
> Redhat and/or Mandrake, I found the automatic hardware detection
> terribly annoying.
>
> > I just know someone has to be working on this, because it seems like such
> > a worthwhile project to enable Linux to escape the hardware headache
> > monica it has too readly acuired.
>
> I suspect Redhat's automatic detection programs annoyed more experienced
> linux users (i.e. admins) than it helped; most admins are able to run
> lspci in order to see what kernel configuration/modules they'll need.
>
>
> If you really want to do something to help the "hardware headache" (I'm
> not sure what 'monica' means - I couldn't find it in any dictionary) -
> help the linux kernel developers to actually support more hardware more
> stably (like Serial ATA, ACPI suspend-to-ram, etc.)  The "hardware
> headache" doesn't come from having trouble identifying cards, it comes
> from having to be nearly on the bleeding edge of kernel development for
> support of any hardware less than a year old.
>
>
> --
> -- Jason Rhinelander
> -- Gossamer Threads, Inc.
>
>
> --
> gentoo-dev@gentoo.org mailing list

-- 
Bibliotech Ltd.



--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Drivers/Portage

[Chris Bainbridge]

On Monday 19 July 2004 09:29, Martin Owens wrote:
> I think I ort to make myself a little clearer.
>
> it's not about auto detection / auto configuration. it's about information
> gathering and resources, sure auto configuration is an option at a much
> later stage but I'd much prefere a system that provides information to the
> usrs as a primary goal.
>
> The most important part is creating/generating unique id's for hardware
> that you can match the information up to. then if the system gets it wrong
> or dosn't have any information allowing the users to input new data
> (subject to review) which can then be used to expand and revise the
> information.

Gathering information in one place is imho the most useful thing. A livecd 
that boots, autodetects the hardware, looks it up all up on an online 
database with a web page for each bit of hardware where people discuss their 
successes or failures under linux and assign a compatibility rating, would be 
a very useful thing. I've wasted hours tracking down drivers for all my 
unusual hardware (game console stuff, 802.11 & bluetooth cards, cellphone, 
camera that isn't a usbmassstorage device etc.) and I'm sure I'm not the only 
one. Half the time when I buy accessories I have no idea whether they'll work 
under linux. 

You'll get plenty of people submitting info if its an open project (ie. they 
can get the info back, its not just to drive hits to your web site), and it 
works well.

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Drivers/Portage

[Hasan Khalil]

Me too! Me too! (See below for inline responses).
	-Hasan

Martin Owens wrote:

> it's not about auto detection / auto configuration. it's about information 
> gathering and resources

I wholeheartedly agree with this -- setting up a system that would tell 
us about our hardware automatically would be sweet. Whether we use that 
system to later then autoconfigure it is a different story, and should 
obviously be a disable-able feature. But what's important is me being 
able to plug something in, and finding out right away if anyone else has 
ever been able to get the hardware to work under Linux <my version 
here>, or at all under any Linux version, perhaps with user comments on 
how they got it to work. I envision this much like a wiki (I don't mean 
to say that a wiki is the answer!) on which our 'autodetect' program 
looks up information for a particular piece of hardware.

> The most important part is creating/generating unique id's for hardware that 
> you can match the information up to. 

Doesn't this more or less exist with pretty much any hardware interface 
(PCI vendor/device id's come to mind for PCI, etc.)?

--
gentoo-dev@gentoo.org mailing list