Re: OT] Joerg Schilling flames Linux on his Blog

["Måns Rullgård" <mru@inprovide.com>]

Chris Friesen <cfriesen@nortel.com> writes:

> Joerg Schilling wrote:
>
>> There is still no new and definitely stable interface that allows me to
>> asume that it makes sense to put effort in implementing support for it.
>
> Why not just accept *any* device node that the user passes in?

That's the question I think we're all asking.  The answer is that
cdrecord *already* does this.

In my understanding there are three basic parts involved here: 1) SCSI
commands, and the devices that use them, 2) low-level
addressing/transport used by the OS to deliver SCSI commands to the
devices, and 3) high-level addressing used by userland when talking to
the OS.  2 can be more or less anything: traditional b/t/l SCSI,
ATAPI, parport, USB, etc, each with it's own unique addressing style.
3 is what applications should be using, and varies across OSes, Unix
systems using device nodes, others using other methods I do not know.
Through a thorough misunderstanding of the word "portable",
Mr. Schilling appears to be attempting to forcibly extend one
particular incarnation of 2 to also cover 3.

FWIW, I run cdrecord using dev=/dev/cdrw as a regular user, without
any suid bits, and close my eyes while the warnings scroll by.  Using
a small window and the -v flag, they're out of sight before you know
it.  I have yet to burn a coaster not resulting from bad media or
hardware, even on loaded systems, over NFS, or whatnot.

-- 
Måns Rullgård
mru@inprovide.com