94-178R0 -- Proposal for TEST SUPPORT command

Ralph Weber -- VMS -- ZKO3-4/U14 weber at star.enet.dec.com
Thu Sep 8 14:02:28 PDT 1994


Gene Milligan wrote:

> In principal I have no objection to some form of the command providing it 
> is made optional rather than mandatory. A device which implements only
> mandatory commands should not be automatically burdened with this command.
> Has the maximum memory size for this command been calculated? Is it intended
> that an issue of the command retrieves all information for all possible
> commands allowing a one time only (repeated as often as you like) mapping 
> of the response? Or is it intended that a nearly limitless combination of
> commands be parsable in any conceivable order?
 
To address the issues individually:

> In principal I have no objection to some form of the command providing it 
> is made optional rather than mandatory. A device which implements only
> mandatory commands should not be automatically burdened with this command.

I believe that this argument is specious.  For example, according to SCSI-2 
Rev. 10K, a disk that implements only mandatory commands cannot have data 
written to it.  Furthermore, the TEST SUPPORT command will provide equivalent 
support for was a requirement in SCSI-2 (the target's testing of reserved 
fields for zeros).  Thus, I think the TEST SUPPORT command must be mandatory.

> Has the maximum memory size for this command been calculated? 

If the statements of those who removed the requirement for testing reserved 
fields are to be believed, then the memory requirements produced by this 
proposal should be insignificant when compared to the savings in code space 
|from the eliminated tests.

> Is it intended that an issue of the command retrieves all information for
> all possible commands allowing a one time only (repeated as often as you
> like) mapping of the response?

One TEST SUPPORT command returns information about a single operation code.  
The return data format does not include the operation code values.  So, 
returning information about multiple commands is not possible (with the 
current data format).

> Or is it intended that a nearly limitless combination of commands be
> parsable in any conceivable order?

I am having some difficult parsing this question.  However, I suspect that the 
answer is, "yes."  Certainly, I envision the following sequence of commands as 
being valid: INQUIRY, MODE SENSE, TEST SUPPORT, READ, TEST SUPPORT, WRITE ...

I guess the klinker is CHANGE DEFINITION.  But, that's always true.

Ralph::




More information about the T10 mailing list