HCL Considered Harmfull [Re: Fedora HCL guide writers?]

[Paul Nasrat]

On Sun, 2005-06-05 at 18:40 -0500, Tommy Reynolds wrote:
> Uttered Patrick Barnes <nman64 at n-man.com>, spake thus:
> 
> > Chidananda Jayakeerti wrote:
> > >Is it possible to achieve an HCL for limited set of hardware, to begin
> > >with. Servers and workstations can be a good start.  We do not have to
> > >maintain an exhaustive list of webcams' keyboards, printers
> > >(linuxprinting does a good job)  and mice etc.
> > >We could also have several maintainers for the HCL based on categories.
> > Trying, especially with a non-commercial project, to maintain any sort
> > of HCL is a dangerous effort.  Those writing the HCL will always be
> > chasing the facts.  Things simply change too rapidly.  
> 
> Well, that may be overstating things a bit.
> 
> Consider: every X months (X=3 or so), a new FCn release is made with
> a FINITE number of included device drivers.  Currently we are in the
> semi-ridiculous position of attempting to offer an operating system
> without even an un-clear statement of what hardware that release 
> supports.

I really think you completely underestimate the task just a single
example - wireless cards are a particular bane as a manufacturers card
will just change chipset with no other necessary difference.  So a
vendor/brandname wireless card can go from supported to unsupported
depending on what "version" (usually invisible) you purchase.  Sometimes
you can differenciate between things like Made in X or Made in Y on the
back of the card, but without full chipset info on the boxes having it's
fruitless.

> Now, there are actually only a handful or so of new devices added
> between FC(n) and FC(n+1).  

But things change rapidly - eg support for synaptics/input devices does
vary between kernel versions and also laptop models (some people seem to
have quite odd issues) - again you need to have documented the touchpad
chipset (ALPS/Synaptics) and against kernel, xorg driver and particular
laptop quirks.

> After the initial effort of producing the
> first:
> 
> 	Manufacturer,Device,(FCn version, tested date)*

See above why this is impossible for some hardware to provide the user
with any particularly meaningful data.

Paul