Should SIP require detection of duplicate tags?

Ralph Weber -- VMS -- ZKO3-4/U14 weber at star.enet.dec.com
Sat Apr 16 14:23:43 PDT 1994


As I recall, one of the arguments against requiring duplicate tag checking on 
the serial protocols was based on SBP.  The SBP design essentially says that 
the tag is the address of the message block containing the CDB.  Thus, under 
SBP a host that sends a duplicate tag also has corrupted the CDB and associ- 
ated data for some previously issued command.  It was argued that, in light of 
the severity of the related problem, duplicate tag checking made no sense for 
SBP.

As for my opinion of duplicate tag checking . . .  I really don't care whether 
the target checks for duplicate tags or not.  However, I place more value on a 
target that does not hang or corrupt unrelated operations when presented with 
duplicate tags.  This value is even more important in multi-initiator systems.

Why should one sick host kill a drive that is servicing many hosts?  The 
method by which a target protects itself from stupid host tricks is not 
important.  That a drive protects itself, is important.

Thanks.

Ralph::




More information about the T10 mailing list