[Info-vax] What is a "real" Unix ?

[Stephen Hoffman]

On 2023-09-04 12:15:22 +0000, Simon Clubley said:

> On 2023-09-02, Bob Eager <news0009 at eager.cx> wrote:
>> On Sat, 02 Sep 2023 23:07:24 +0200, Johnny Billquist wrote:
>> 
>>> On 2023-09-01 14:50, candycane wrote:
>>>> SC> Yes, at least on Linux.
>>>> 
>>>> SC> The dd on other operating systems may use a different signal.
>>>> 
>>>> Is dd on non unix systems?
>>> 
>>> Yes. Linux for example. Linux is not Unix, but it certainly tries to
>>> work the same.
>> 
>> "Jumped uo UNIX wannabe"
> 
> In that case, what is a "real" Unix ?
> 
> Is it something that implements a set of user-visible APIs and certain
> behaviour within its kernel (fork() semantics for example) ?
> 
> Is it something that came from a specific source code base and hence
> nothing else can never be called Unix no matter how compatible that other
> something is ?
> 
> If BSD is a Unix, then is System V also a Unix ?
> 
> If System V is a Unix, then why can't something else that also implements
> the same APIs and kernel behaviour also be a Unix ?
> 
> Or is Linux really a Unix after all (in every way that matters) and what's
> really going on here is just some out-of-touch BSD Unix elitism ?
> 
> Simon.

Here are the vendors and products that that have certified and that 
accordingly have permission to use the UNIX® trademark: 
https://www.opengroup.org/openbrand/register/

Product sales and marketing organizations can be interested in passing 
qualification or certification tests and licensing trademarks, when the 
buyers are including those tests or branding requirements in their RFQs.

OpenVMS was within range of passing the then-current compatibility 
qualification testing an aeon or so ago with the DII COE work 
(V7.2-6C1, V7.2-6C2), as that was effectively then a test for Sun 
Solaris compatibility. Effectively, the US Government wanted to buy Sun 
Solaris servers and software for some then-current COTS-related 
projects, but couldn't legally specify Sun hardware and software in the 
RFQ. The Single UNIX Specification and The Open Group were all getting 
going around that same Y2K-ish timeframe.

Here is an actually-still-active HPE OpenVMS link on DII COE: 
http://h41379.www4.hpe.com/solutions/government/coe/certification_programs.html 





Unrelated, and entirely for what it is worth... While checking some 
details and dates on DII COE availability here—we're further from Y2K 
now than Y2K was from the VAX/VMS announcement back in 1977—I found a 
nice little write-up on perceived (and measured) I/O performance issues 
from a Y2K-era thread comparing OpenVMS with Linux. A similar 
discussion is happening around here again over in another thread. The 
following quote is from Y2K, and is by David Mathog:

"For a UNIT text file operation, going RAMDISK to RAMDISK on OpenVMS, 
or file cache to file cache on Linux, on identical DS10s, the result is 
that the Linux system is 2.5-6.5 times faster. This is for an operation 
like "read text record from input, write text record to output" - pure 
IO. It doesn't matter if you do this in 1 file or in 1000 you're 
already starting out with the Unix systems "lighter" text file handling 
mechanisms 3X faster than those on OpenVMS. And it goes downhill from 
there, rapidly, because the lack of file effective file caching on 
OpenVMS throws in another factor of 10 advantage to Linux. The only 
time I can get similar performance from my VMS box is when the IO is 
minimal (a CPU bound program) or the IO is done differently, via memory 
mapping or some other mechanism to bypass RMS.)
https://groups.google.com/g/comp.os.vms/c/metuuEsFXLY/m/9o3ODbwJAPQJ






-- 
Pure Personal Opinion | HoffmanLabs LLC