On Thu, 16 Nov 1995, Brett Lymn wrote:

> Dunno if this has been said before but what is wrong with just using
> sdn where n is the SCSI id of the device? [...]

This is not very general; especially when you consider folks like me, who
(for a very good but basically irrelevant reason) have their first disk at
target 0 sometimes, but target N!=0 most of the time.  Either disk will
boot, but it is very reassuring to know that I can use the same GENERIC
kernel to operate on both. 

Also, it unnecessarily complicates installations by requiring someone or
something to find out what disk to install on.  If you write user
interfaces, you probably know the prime assumption:  "users are idiots." 
(Lest I be misinterpreted, I don't think any individual reading this is an
idiot... but statistically speaking, there is a non-zero chance that I'm
wrong and that some "mentally challenged individual" actually _is_ reading
this.  No offense intended.)

> The worst case scenario I can think of
> off the top of my head is that you have /usr on a separate disk and
> you lose a disk in the scsi chain, suddenly your /usr is not mounted.

It has already been mentioned that if you have this concern, you can 
build a kernel with devices mapped to specific SCSI targets/LUNs.  In 
fact, you should probably do this anyway, to get rid of unused 
drivers/emulations/etc.  If you want to live under the limitations 
imposed by a GENERIC kernel (for your definition of "limitations"), it's 
entirely your own choice.

--scott