>Below is a draft request. In fact damfino why we should ask for a long
>word that is in range 0-255.Might as well be a byte if Steve prefers.

It was Steve that suggested a longword.  Probably in case we ever want to
expand, or use a byte or two for something else, but mostly, I suspect, to
preserve longword alignment.

>This unless of course we want to allow a larger value later. In that
>case why not ask for a word or longword but document that it should be
>0-255 for now and have Steve be sure it's value is 99999 or less (so
>number representation doesn't bite)? That's the only "hard" limit I know
>of for these things. MSCP server will limit to 32767 moreover, but
>there are cases where that won't matter.

I've never looked, but I would have guessed that the allo class could be
999999, since the SCS node name that it replaces in the device name is six
characters.  Why did you pick 99999?

>I agree that documenting 0-255 and using that is a good idea, but firmware
>is likely to be relatively immovable and if we want ever, even conceivably,
>to allow more than 0-255 we ought to say so.

Okay, how about if we ask for a longword, and reserve two bytes.  That
gives us an allocation class of 0-65535.  The firmware rejects values
outside this range, and the documentation says to use values in the range
0-255.  Make sense?

>To: Steve Sicola
>From: Glenn Everhart
>Re: Controller allocation class in HSZ INQUIRY data
>Date: 11-March-1997

>Steve -

>In order to provide naming uniqueness for HSZ devices with shared
>paths, we would like to have an allocation class return from the
>devices themselves, which can be set by the customer. 

Better be more explicit:  ...an allocation class returned in the inquiry data
from each virtual device.  The allocation class value will be set by the
customer at the HS* console.  This value will be non-volatile and the value
will be returned by all virtual devices on either controller in the pair.  

>The range
>0-255 is adequate for this allocation class, where we can take 0
>to mean none is needed.

>It should be returned somewhere in the INQUIRY data for the HSZ.
>Makes no difference to us where, just so we know and so it is constant.