reddred wrote:
"Scott Dorsey" wrote in message

Now it's true that the v7 kernal, the BSD kernal, and the SysV kernal are
radically different from one another, but they all have supersets of the
same interface to the applications code, which makes the all Unix.


I have trouble accepting that this issue boils down to interface
compatability, when the philosophies of all of the above are indeed
radically different. And there are obviously legal differences. If an OS
meets the proprietary spec, it doesn't mean they want to be called 'Unix'.

That has become the difference between "Unix" and other things. Unix supplies
a "POSIX-Compliant" (whatever that really means now) interface between the
kernal and the appliction level.

Mach and VMS don't have the same interfaces from the kernal at all, even
though they have compatibility libraries which allow you to make those
system calls to an intermediate layer and have them work.


Which is probably a good thing, I agree. But the main difference here,
technically is microkernal vs. monolithic. That's something that still has a
decent technological definition. Unix doesn't anymore.

No. VMS is a huge monolithic lump. It has an additional layer of junk
between the kernal and the application available in order to fake a POSIX
application interface. It's not a microkernal, but it's not Unix either.

I guess what I was getting at is, what are people supposed to call Unix-like
systems?
'Loose-collection-of-mostly-posix-compliant-bits-of-operating-systems-that-a
ct-like-unix-but-arent'?

Unix-like. Xinu is Unix-like. Minix is Unix-like.

Zdnet will never print an article called 'Apple - keeping Mach alive' but
they will call MacOS 'Unix' over and over. The integration at kernel-level
of Apple's Mach with BSD clouds the issue even further. I guess it's not
written in stone, but I'm not sure where the motivation would come from to
change it at this point, so for all practical purposes it remains
'Unix-like'.

Unix-like is okay. But saying it's Unix isn't to my mind.

and reminiscent of some
of the nasty kludges used to make VMS and Unix machines talk.

Got any horror stories?

Lots and lots. The thing is that VMS has a dual-fork filesystem as well
if you want to think about it that way. There is formatting information
stored in the directory along with each file. So you can have a text file
that is a STREAM_LF file, or is a variable-record-length file with 80 column
maximum length records, and they seem the same from the user perspective but
they are different.

This is because the operating system itself has a lot of database stuff
built into it, so you can just open up a file and say it's an ISAM or KSAM
database file and all of the database stuff gets done for you.

But now when you NFS mount that file from a Unix box, all of the information
is lost. You take that ISAM file and use ftp to copy it over to another
machine, and all of a sudden the OS thinks it's a STREAM_LF file and you
need to go in and tweak the format parameters by hand to make them match.

The VMS filesystem is really amazing for commercial database applications,
and it's really a shame to see it going away in favor of lowest common
denominator heirarchical filesystems like the 4.2 and NT filesystems.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."