Subject: Re: 32 bit dev_t, Revision 2
To: <>
From: Henry W Miller <mill0440@gold.tc.umn.edu>
List: tech-kern
Date: 01/12/1998 10:53:17
On Mon, 12 Jan 1998, Manuel Bouyer wrote:

> On Jan 11, Chris G. Demetriou wrote
> > 
> > 8 bits to select a SCSI bus, 4 to select a target, 3 to select at LUN,
> > 5 to select a parptition.  8-)

> Is 16 targets/bus enougth for future SCSI and SCSI-like standarts ?

Not with the above 4-3-5 scheme.  fibrechannel supports 100+ (128 I'd
guess, I can't recall offhand) devices on an arbritrated loop, and I will
have to work with such a large configuration.  with only 4 bits to select
the device on the bus that won't work.  (accually fibre channel supports
over 1000 devices, but not on an abritrated loop, my boss has told me
abritrated loops are the way to go so I havn't bothered looking at the
rest)

In fact the (preliminary, it isn't released yet) SCSI-3 standard already
defines a 32 bit bus, supporting up to 32 devices on a cable.  There is
currently a footnote that explains current cabeling technilogy doesn't
make this pratical now.  The standard is defined however, and I would
expect the cabeling/signal problems to be worked out in the future.  If
they are not worked out it means that fibre-channel took over.

I would not recomend limiting ourselves as the above scheme lists, unless
the upgrade to support more deices in the future is painless.

--
       |\      _,,,---,,_  http://www.tc.umn.edu/nlhome/g639/mill0440/    
 ZZZzz /,`.-'`'    -.  ;-;;,_   |  Isn't there always a cat    | Henry
      |,4-  ) )-,_. ,\ (  `'-'  |  on whatever you're reading? | Miller
     '---''(_/--'  `-'\_)  mill0440@gold.tc.umn.edu