Re: Is Unix security really so weak?

In article <5i41io$qrd_at_inet-nntp-gw-1.us.oracle.com>, Bill Manry <B.Manry_at_us.oracle.com> wrote:
>Andrew Gierth <andrew_at_erlenstar.demon.co.uk> wrote:
>>Perhaps you (or someone from the Oracle group) could provide some actual
>>information rather that unsupported statements?
>
>In general, Oracle I/O is implemented on each platform in such a way
>that completion of a write (as viewed by the software) means the data
>has been fully recorded on the external medium. In some environments
>there are mechanisms which seem to work (NFS-mounted files, for
>example) but which are expressly not supported by Oracle because they
>do not meet this requirement. The platform-specific documentation is
>responsible for indicating what is or isn't supported.

Yes, precisely. I knew that, which is why I posted what I did (which was NOT quite the same as what I was quoted as posting).

>I'm not aware of any systems where Oracle is capable of using the
>normal "cooked" I/O facilities with real, local disks but it fails to
>meet the above requirement. But perhaps the earlier poster can point
>one out...I'm no UNIX authority.

When I last looked at the area (some 5 years ago), perhaps 25% of the main Unix systems either had no fsync() or documented it as being unreliable, even for local file systems. More seriously, some fairly simple tests indicated that about another 25% had it, but it didn't work AT ALL. On the remaining ones, I could not prove it either way. I have some grounds for believing that things have improved since.

Furthermore, Oracle implementations can document only what they can find out. Unix man pages are really an aide-memoire rather than a specification and most of them (on most systems) have serious omissions or errors; and this is exactly the sort of thing where they are unreliable. It is unfortunate that the first reliable specification of most of the Unix calls was POSIX!

Nick Maclaren,
University of Cambridge Computer Laboratory, New Museums Site, Pembroke Street, Cambridge CB2 3QG, England. Email: nmm1_at_cam.ac.uk
Tel.: +44 1223 334761 Fax: +44 1223 334679 Received on Sat Apr 05 1997 - 00:00:00 CST