This message is from the T13 list server.
[ BC [EMAIL PROTECTED] ] // Everyone: > Could Serial ATAPI force the standard set of > ATAPI calls ? I leave that question open. // Gilles M: > > In the Standard INQUIRY data > > (for the SCSI INQUIRY command), > > bytes 58-73 are defined as VERSION DESCRIPTOR fields > > that can be used > > to indicate the versions of the standards > > with which the device complies including MMC version. > > I admit that ATAPI devices theorically provides > this version descriptor. Unfortunately some > recorder drives simply never fills up ... Eh? The reality I see departs further from the t10 fiction than that. Me, I see hosts massively distributed in the real world that choose not to risk any op x12 Inquiry other than the -i x24 -y "12 00:00:00 24 00", because anything other than that standard initial Windows Inquiry surprises a significant minority of potentially connected devices beyond recovery. Consequently, paper talk of what bytes may exist past offset x24 concerns only effectively vendor-specific arrangements, such as the painfully finite list of supported drive names that appear in dvd/cd recording software. > Unfortunately some recorder drives simply > never fills up the REQUEST SENSE (0x03) > request fields, ... Help! I'm lost. How did we slip over to talking about op x03 RequestSense rather than op x12 Inquiry? > Unfortunately some recorder drives simply > never fills up the REQUEST SENSE (0x03) > request fields, it just fills the error code > (0x70 or 0x71) and no sense key. This behavior > is written nowhere. Ouch. Thanks for sharing this history of pain. Can you elaborate? What evidence persuades us that some drives fail without offering the host its one transient chance to read-and-clear nonzero SK and/or ASC and/or ASCQ via op x03 Request Sense? This rumour surprises me because before now I had heard that every host less reliable than an Apple Mac was less reliable in part because those hosts by design chose to pretend they knew what the SK:ASC:ASCQ meant, as a consequence of having chosen to make the drive visible to the GUI/CLI even when no volume is mounted. Part of pretending we know what SK:ASC:ASCQ mean naturally includes assuming the drive bothers to copy SK:ASC:ASCQ into us on request. Pat LaVarre P.S. I can see auto sense is headed for trouble. Win ME/9X asked for x0E bytes, Win XP/2K asks for x12 bytes ... and neglecting to align address and length to sizeof (int) is by definition a way of asking for trouble. I expect choking over length x12, aligned to 16 bits, to hit us before choking over length x24, aligned to 32 bits, does.
