Subject: Re: Chip vs. card compatibility
To: None <netbsd-users@netbsd.org>
From: Andrew Smallshaw <andrews@sdf.lonestar.org>
List: netbsd-users
Date: 04/18/2006 21:40:51
On Sun, Apr 16, 2006 at 11:16:43AM +0100, Dieter wrote:
> >   3) The database is both "chip" and "baord" based, and contains a
> > table which associates boards with chips.
> 
> Yes, having both chip and board names would be very useful.
> 
> For board, include devices such as the external boxes with
> Firewire and/or USB to PATA/SATA drives, adapter/bridge cables
> such as USB-to-Ethernet, USB-to-RS232, SCSI-to-SATA, and so on.
> 
> I think I remember reading about cases where a board changes
> the chip with changing the name of the board.  The database
> should have a provision for alerting users to this.

There are also cases where the same chip is used in mutually incompatible
cards.  The only example I can think of concerns the 3com 3c905.  Under
SCO OpenServer which I was using at the time the 3c905 driver would
refuse to play with the 3c905B when it was new out, until the driver was
updated, even though both use the same chip.  Doubtless there are other
examples but I can't immediately think of any at the moment.

The chip vs card debate depends essentially on what hardware is being
considered.  Well known brand-name hardware could be listed with 
relatively few caveats which can be addressed in free text fields.
Cheap no-name commodity hardware really needs to be addressed on a per-chip
basis if the database is to remain remotely managable.

IMHO it comes down to what question is being asked: Is it "will my
computer run NetBSD?" or "What hardware do I need to specify for this
machine which will be running NetBSD?"  The former needs a comprehensive
listing of all computer hardware out there - this is impossible so can
only really be considered an aid rather than authoritive.  The second can
be done effectively by selecting a few known-trouble-free devices in each
category (though not all high-end, nor not all cheap tat).

In summary I'd suggest a combination approach -it'll never be all things to
all people but that would seem to cover the most bases.

-- 
Andrew Smallshaw
andrews@sdf.lonestar.org