Joerg Schilling <[EMAIL PROTECTED]> writes: > - Trying to work around a Linux design problem that has been introduced > about 3 years ago when Linux stopped to support an orthogonal SCSI > Kernel transport for all SCSI devices.
This is nonsense. Fact is that /dev/hd* ATAPI devices support the same SG_IO passthrough as /dev/sg* SCSI devices. All barriers that make ATAPI look different than SCSI are artificially created inside libscg, and if you want numbers, the linux-kernel archive has sufficient information on how to obtain them, as proven by the patch I sent to linux-kernel in early 2006. (Which brought nothing but personal offenses, Jörg's claiming - without proof, of course - my patch made matters worse.) > Since then, ATAPI drives are handled different from other SCSI > devices. Gosh, they added /dev/hd* to the bill. That's the end of the world =:-> > The problem is that the linux maintainers intentionally reduce the > information that is available in the kernel and this way prevent to > allow libscg to only show only unique drives. You have a SINGLE handle to access the device as block or as character device, for SG_IO and everything -- that's something /dev/sg* still can't do, if you want cooked access, you're still guessing which /dev/sr* might match your /dev/sg... ATAPI doesn't have that problem. > libscg now tries to map ATAPI drives to SCSI bus numbers >= > 1000. What a botch. Perhaps it's time that cdrecord learns to treat device identifiers as opaque string, without second-guessing a numbering that falls apart the moment hotplug comes into play. > J\366rg Still not fixed your mailer? -- Matthias Andree
