Eduard,
if you have no clue about SCSI and kernel design, please do not comment
things you do not understand.
>> Using something like "dev=ATA:2,0,0" still gives you
>> SCSI emulation, right?
There is not SCSI emulation at all!
Read README.ATAPI:
....
Well first a statement: There is no single IDE burner out!
Even a CD-ROM cannot be used decently if you use only IDE commands.
Opening/closing the door, playing audio and similar things
cannot be done using vanilla IDE commands - you will need SCSI commands
to do this. But how do we do this with a drive that uses an IDE
interface?
ATAPI stands for ATA Packet Interface
The ATAPI standard describes method of sending SCSI commands over IDE
transport with some small limitations to the "real" SCSI standard.
SCSI commands are send via IDE transport using the 'ATA packet'
command. There is no SCSI emulation - ATAPI drives include native
SCSI command support. For this reason, sending SCSI commands to ATAPI
drives is the native method of supporting ATAPI devices. Just imagine
that IDE is one of many SCSI low level transport mechanisms.
The module ide-scsi on Linux just acts as a SCSI host adaptor driver
for ATA transport for SCSI.
>Wrong. It is just our bullheaded upstream who tries to force his name
>convention over every device type, even where it makes absolutely no
>sense.
It seems that the right answer for your drivel is to tell you that you
seem to be driven by the same unwillinglness to self educate
as e.g. Linus Torvalds.
So my advise to you is: educate yousdelf, read the SCSI standards, check
other OS implementations and find out that what the Linux kernel authors try
is just a silly and childish attempt to prevent users of cleanly written
software from doing what they like to do.
>> According to kernel/Documentation for 2.6 it should be possible to
>> access the IDE device directly, without additional onion layers.
There was never a need to introduce additional layers. However,
Linux 2.6 introduces so many different and unneeded approaches to send
SCSI commands to devices that following my advise could reduce the kernel
size by more than 200 KB.
The right and most lightweight solution would be to write decent SCSI
kernel drivers, a decent and generic DMA setup for kernel drivers
and treat ATAPI devices like they have been designed: as SCSI devices
that use ATA as a transport. To allow this, a very small driver is needed
that integrates ATA interfaces into the list of SCSI host adaptors.
Operating systems that go this way are:
SunOS (Solaris)
FreeBSD
Win32
OS/2
BeOS
Zeta
HP-UX
VMS
SCO-OpenServer
SCO-UnixWare
MacOS
If you believe that the majority of OS is wrong, please send a description!
>It was possible in the first implementations then JS changed it to force
>people to use his favorite syntax. Bug him directly, he does not listen
>to sane arguments any more.
As in most cases, you are completely wrong.
>> _This_ should be documented.
>We need a good FAQ explaining such details, written by an unbiased
>person (especially not listening to JS' repeating rants against Linux
>developers).
If you do not like biased text, then please stop your biased drivel
that just verifies that you have no clue.
Read README.ATAPI, it is _the_ FAQ and it exists for a long long time.
Jörg
--
EMail:[EMAIL PROTECTED] (home) Jörg Schilling D-13353 Berlin
[EMAIL PROTECTED] (uni)
[EMAIL PROTECTED] (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
