X-Mailer: exmh version 2.0.2
To: linux-kernel@vger.rutgers.edu, linux-scsi@vger.rutgers.edu
Cc: ulmo@armory.com
Subject: I found SCSI driver ratings by Schilling ranking Linux as "poor" and
"unsuitable for development on"
From: Brad Allen <ulmo@deepthought.armory.com>
Reply-To: Brad Allen <ulmo@deepthought.armory.com>
Date: blah
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
While reading the release notes to CDRecord on
http://www.fokus.gmd.de/nthp/employees/schilling/cdrecord.html, I
found this paragraph:
The user land SCSI transport implementations on Linux and *BSD are
the worst ones. Linux will not tell you all errors while *BSD hides
the device files from you. SGI does not allow SCSI disconnects on
odd byte count boundaries. Look for my SCSI implementation ratings.
His last sentence links to his ratings page:
http://www.fokus.gmd.de/nthp/employees/schilling/scsirating.html
In that page, he ranked Linux as ``poor'' (below ``sufficient'' in his
scale), with a special note that it is ``Not suitable for development
of new programs.'', the worst in the list. BSD was the only one to
fare as bad, with a ``poor'' rating; SGI was ``fair'' (above
``sufficient'' according to his scale), with a note about some bugs.
Sun's Unix OSes (SunOS & Solaris) each garnered a 2 (``good''), as did
HPUX; HPUX would have fared better (``excellent'') if it had included
some device files (I suppose they omitted some mknod's or something -
i.e., packaging). So, according to this persons' opinion, HPUX is
rather excellent with a few small problems, and Sun's OSs are getting
there, whereas the rest are not.
See the page above to fill in the rest of what he said, although I
find it reasonable to include his final paragraph(s) for rounding out
this note:
I am sad to see that the two free UNIX clones have the worst SCSI
user level transport implementations.
NOTE: The ratings mainly apply to the usability of the SCSI user
level transport mechanism. They have been maded when porting cdrecord
to the various architectures. For this reason, it does only include
known problems of the overall behavior. The judgement for the whole
SCSI implentation may be different if you spend more time with
testing then I did.
So, I approach the 2.1 era and say, ``what have you had a chance to do
with regards to this problem?'' Hopefully, this is already all fixed
(I have steered away from development kernels for some years now so I
do not know), or plans are well underway for 2.3, but for all I know,
the reality could be that this fire needs stoking, or, even, lighting.
So I submit to you.
Sincerely and thank you all very much for enjoying and improving the
same OS I take refuge in as well,
-u
Brad Allen
<ulmo@deepthought.armory.com>
(temporary email address during the dark ages of dynamic IP)
--==_Exmh_386013222P
Content-Type: application/pgp-signature
-----BEGIN PGP MESSAGE-----
Version: 2.6.3ia
iQEVAwUBNXuIRZxWhFYc6x9VAQEleAf8DB5Q63HyEWP/9arjD9aaUV2mG8n3xWqe
Kiw66dJ4rU2I9eeAQbk1e4yZ3h54wL8ENSLTCBm0l3soluWBWsVA0w21H2XklOee
d1xziUGMHnNPIiooRAF6xIu38lbA02AQajGzSU6L3srneg21kSjwajftY3sGI/p2
L56DSH4mtnmodw96EYWxoyE+nMJz1pxC9cYomrQmOJ7p5jNKYlpI5IutR2B553TQ
4d391PUsNoRNmW0wJL0yuRiK/M0wvXvH71uYPcdWlL/U6ruqsXJupDBt+Y1yHh1L
IPZoKCXrbr4HLnDGqNtlL4PGYsx4csO5b4hHlISZh4/UoEUpltcymg==
=D5Cw
-----END PGP MESSAGE-----
--==_Exmh_386013222P--
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu