[Info-vax] Current VMS Usage Survey

[johnwallace4 at yahoo.co.uk]

On Monday, 2 December 2013 05:36:34 UTC, Subcommandante XDelta  wrote:
> (Was: Re: READ and WRITE vs. SEARCH/OUTPUT)
> 
> 
> 
> Subcommandante XDelta:
> 
> > On Wed, 8 Feb 2012 03:46:45 -0800 (PST), Keith Cayemberg
> 
> > <keith.cayemberg at arcor.de> wrote:
> 
> > :
> 
> > :
> 
> >> There are many examples of OpenVMS being used for Real-Time
> 
> >> applications, especially apps that required resources beyond the
> 
> >> capabilities of embedded systems at the time. To my knowledge the
> 
> >> Flight Control System governing the entire North Atlantic airspace and
> 
> >> based in Iceland is still running on OpenVMS. There are also several
> 
> >> radar systems, military communication systems and satellite control
> 
> >> systems running on OpenVMS. 
> 
> > 
> 
> > May we assume that some of the VMS userbase, at least, would have
> 
> > strong opinions about the extinction of VMS if they knew it was on the
> 
> > drawing board?
> 
> 
> 
> Mr Mezel, in a recent conversation thread, estimated, IIRC (senior
> 
> moments notwithstanding), that the remaining VMS end-user population
> 
> might be circa 100,000 shops.
> 
> 
> 
> So this topic is not about who or what those shops are, but rather what
> 
> those shops are using VMS for.
> 
> 
> 
> Port the FCS cited above to NSK from VMS, with new generations of
> 
> Bohrbugs and Heisenbugs to iron out? - not bloody likely.

There's another reason why NSK is irrelevant to some VMS systems.

NSK is fundamentally a "transactional" view of the world: start transaction, 
read some stuff, do some magick, write updated values back(?), end transaction. 
Transactions succeed or fail atomically (just as they do on any other 
conceptually similar database system). That's the high level model.

VMS can do the transaction stuff, especially when accompanied by databases, 
RTR, all that good stuff.

VMS can also do multiple other classes of application that don't really do 
transactions in the NSK way. Whether these applications *could* be redesigned 
to do that is currently irrelevant, currently they use shared memory, networks, 
ASTs, and the rich runtime environment that allows folk to run both the 
realtime control and the database stuff on one pair of (failsafe, loadsharing, 
not necessarily clustered) boxes. Moving such a long-lifetime setup to NSK 
isn't going to happen: (1) the programming model isn't appropriate for a 
cost-effective and trustworthy port (2) the current OS owner isn't trustworthy 
as the long term owner of an essential component of your business.

There's trivia like XD Ada in this picture too - that's not going to work on 
NSK. Yes there's gcc/Gnat, but when you have a flight certified setup, 
re-certifying at some vendor's (ie HP's) whim isn't going to win the vendor any 
friends.

Fortunately for HP the number of systems in this picture is quite small.

On the other hand, for a business largely focussed on VMS, the number of 
systems involved (and more importantly the money involved) might be worth a 
second look.

In much the same way as a VW car dealer network wouldn't be interested in 
selling light commercial vehicles because they don't understand either the 
light commercial vehicle products or the markets, and in particular the 
differences between the two. But sell the identical light commercial vehicle in 
somewhere that does understand the market, and suddenly it's a commercial 
success again. (thanks to, if I remember rightly, Paul Sture for fleshing out 
my cars vs minibuses 'hypothetical' with an actual real world recent history 
example - sorry Paul, Google Groups search has been 'improved' so much I can't 
quickly find the post in question).

Have a lot of fun.