discussing flavors of POSIX [was M$ && Sending files back?]

[bogstad at pobox.com (Bill Bogstad)]

Jerry wrote:
>Bill,
>Neither does BSD. BSD does not contain any code from the AT&T kernel. 
>Additionally, there are many Unix systems (based on OSF1 for instance) that 
>also do not contain AT&T code, but they are branded Unix. 
>Certainly, System V was pure Unix, and the Berkeley Unix releases were 
>originally based on Unix version 6 (which predated System V). 
>I don't have the time to go into the dates, but back in the early 90s, when 
>AT&T sued BSDI, all AT&T code was removed from the BSD kernels. Both Clem 
>Cole, in his talk last year presented a roadmap outlining it. 

	Yup, I know that.  My first Unix usage in 1983 was on a PDP-11
using a modified v.7 UNIX.  Then I moved to a VAX runing 4.1c BSD.  Of
course, this was all back when BSD DID contain AT&T UNIX code.  Even
when confronted with the lack of AT&T UNIX code in current BSDs, some
people (usually on USENET) continue to claim that somehow BSD is more
Unix then Linux because many of the people involved in continuing *BSD
development at some point had access to AT&T source.  It's really
bizarre.  I get the same feeling when I read discussions where people
say that Linux (it always seems to be Linux) could NEVER be as good as
Solaris.  Apparently because it's not blessed by Scott McNealy.  Of
course, even that's changing...

>While Linux was created from scratch, I personally include it when I talk 
>about Unix, and it does comply with some POSIX standards, and interops with 
>most Unixes.

	The same people sometimes describe Linux as being based on
Minix with the strong implication that the source code was copied
over.

> There was a time when FreeBSD was clearly a better OS that 
>Linux. Whether that is true or not today, I don't know. Linux has come a 
>long way, and in some vendors, it is beginning to replace their commercial 
>versions of Unix. I suspect that if a vendor wanted to spend enough money, 
>they could get away with branding Linux and Unix. (Note the The Open Group 
>owns the Unix brand, and their standards determine whether or not a vendor 
>can use the Unix brand, but Finnbarr is the expert on this, not me). 
>
>I know that even some proprietary systems that were never ever thought of 
>as Unix are now being made compliant with POSIX and even Unix 98. 

Yeah, but will those proprietary systems ever reach the new POSIX
standard forthcoming from the Austin Group activities.  For that
matter, will any system.  Maybe we'll find out at tommorrow's meeting.

On a philosophical note, I find "The Innovator's Dilemna" by Clayton
M. Christensen relevant to the evolving Unix landscape.  He suggests
(and documents) that good companies in markets subject to
technological change spend their time chasing their 'best customers'.
i.e. Those who generate the highest profit margin and who usually
demand the highest end of a particular performance curve.  New
methods/technologies that can't immediately be beneficial to the 'best
customers' are usually given little attention by market leaders even
if they might result in a better product on some other performace
curve.  Such technologies end up being developed by companies that are
falling out of the market anyway or new entrants entirely.  In some
cases, these new technologies can be improved to the extent that even
the 'best customers' find them adequate at which point those customers
desert the market leaders in droves.

If you look at the history of commercial Unix, I suggest that you see
signs of this.  Originally, Unix vendors concentrated on engineering
workstations.  Now they spend most of their effort chasing high-end
server sales.  Linux & *BSD are certainly better on the 'price curve'
and some would argue as good as (or soon will be) commercial Unices on
the other curves.  I see no reason to believe that they can't get good
enough on those other curves to surplant commercial Unix for servers.
Admittedly server OS software selection brings buyin issues with it,
but I believe that recent announcments about major companies using
Linux for serious commercial activities indicate that this is already
starting to happen.

Desktop OSes have even worse buyin issues then server OSes which
Microsoft has managed to use to retain a virtual lock on such systems.
However, even in this area the better 'price curve' is a problem for
Microsoft.  On other curves relevant to desktop users (hardware
support, application availibility, ease of use, widespread
familiarity, etc.), Microsoft can at least suggest that they do better
and sometimes are.  Free Desktop software is getting 'good enough' for
many users and Microsoft software isn't getting any cheaper...

				Bill Bogstad
				bogstad at pobox.com