REQUIRE AN IDENTIFY MESSAGE IN SIP

Tom Wicklund wicklund at intellistor.com
Tue Aug 16 08:36:36 PDT 1994


Duncan Penman writes:
 > 	I agree we've passed a point of no return in the use of Identify 
 > Message Out.  I've no objection to clarifying the expected behavior in 
 > SIP.  I do feel a bit cautious about accepting your request verbatim for 
 > two reasons:
 > 
 > 1) It violates my (admittedly primitive) grasp of SAM's layering 
 > philosophy to specify the behavior of an interface signal in the protocol 
 > document...seems the natural home would be SPI.

If so, then the statement about which messages are accepted should be
moved also.  It seems consistent to keep both in the same place.

 > 2) I can envision a number of other disastrous cases that could result 
 > from broken interface hardware.  My own experience has been that trying 
 > to specifically address any but the most common failure modes is 
 > counterproductive.  I'd like to understand more clearly why this 
 > particular hardware failure should be singled out.

I think this change is intended to remove an ambiguous situation -- if
there is no IDENTIFY message then the target doesn't know which LUN to
associate the command with (since the LUN field of the CDB has been
removed).  The detection hardware failure is a rather unimportant side
effect.





More information about the T10 mailing list