Article 139987 of comp.os.vms:
Path: nntpd.lkg.dec.com!lead.zk3.dec.com!crl.dec.com!crl.dec.com!caen!uwm.edu!vixen.cso.uiuc.edu!howland.reston.ans.net!swrinde!ihnp4.ucsd.edu!mvb.saic.com!info-vax
From: petit_jp@decus.fr (Jean-Pierre PETIT - President SIG VMS)
Newsgroups: comp.os.vms
Subject: A paper on Windows NT / OpenVMS binary compatibility
Message-ID: <96020915434734@decus.fr>
Date: Fri, 9 Feb 1996 15:43:47 +0200
Organization: Info-Vax<==>Comp.Os.Vms Gateway
X-Gateway-Source-Info: Mailing List
Lines: 214

	The following text is a paper on what opportunities Digital
	and the VMS community could take advantage of, given the
	affinity between Windows NT and OpenVMS. It also encompass 
	some vision of what Digital could become or not, depending
	on its strategy.

	Most of the content is based upon feedback from the OpenVMS
	community in France.

--------------------------------------------------------------------------------
	Jean-Pierre PETIT				ESME-Sudria
	DECUS France					4 rue Blaise Desgoffe
	VMS systems SIG chair				75006 PARIS
	petit_jp@decus.fr				FRANCE
							+33 1 45.48.03.70
--------------------------------------------------------------------------------




                 OpenVMS / Windows NT affinity
		 -----------------------------


The need for applications
-------------------------

The end of the 80's has introduced a dramatic shift in the way
investment decisions are made. The availability of
applications and tools is now the major criteria where it used
to be characteristics of the operating system and its
underlying hardware platform. Once chosen for its qualities
OpenVMS is now impaired by the lack of applications, ISVs
being cutting down the cost generated by multi-platform
support.

The Win 32 API
--------------

The Win 32 API on OpenVMS is a unique opportunity to give
OpenVMS users the access to a lot of applications and
utilities. The Windows world has plenty of them in almost any
domain of interest.

Does it solve the problem ?
---------------------------

The Win 32 API is supposed to allow immediate porting of
applications by simply recompiling the source code with no
need to edit it. This sounds good, but lets have a look at the
other sources of induced costs:

1.   The newly compiled code must be tested because some
 hidden bugs may show up on a different architecture.
2.   The binaries must either be included in an existing
package (CD-ROM) or made available in a separate package
(floppy disks) .
3.   The OpenVMS platform must be supported by the ISV's
 support center.

And for all this points to apply, the ISV must be aware of the
existence of the OpenVMS platform and convinced that it does
represent a market large enough to justify the expense of the
port and support. In other terms, whether it would prove to be
profitable or not, the initial step will never get done
because of inertia.

Lets have a look to other APIs and the relationship between
availability of a given API on OpenVMS and the availability of
applications built on top of this API.

First, we had the Xwindows/Motif API. As far as DEC extensions
are not used, any applications written in ANSI C using this
API can be ported on OpenVMS by simply recompiling the source
code. Did that provide the OpenVMS community with a lot of
applications ? The answer is NO, and it looks like
applications ported a few years ago will no longer be
supported.

Then we got the POSIX API. How many ISV's applications did
that bring to the OpenVMS community ? Well, I am still looking
for such a list !

Clearly, the availability of an API on a given platform is in
no way a guaranty for the availability of applications.

Binary compatibility
--------------------

There is a better way to start the process of regaining
applications, and it is named "binary compatibility". The aim
is to be able to use any application without recompiling. What
that means is that ISV's are no longer involved in the initial
start-up of the process since binary compatibility enables
anybody to put any existing software right out of the box on a
new platform.

This is the way for the OpenVMS community to get applications
first and then prove to the ISV's that the market does exists
and that it must be supported. All success stories have
followed this bottom-up approach as opposed to the lots of
failures in trying the other way (top-down approach which
means  trying to impose a standard to a market which don't
care about it). Not yet convinced ?  Think about the "OSI Vs
TCP/IP" case.

Let's examine the two flavors of binary compatibility (I will
not examine software emulation because of its poor
performances).

Binary translation

While it is a new technique, it has proved to be viable in the
VAX to Alpha transition. Digital has the needed expertise to
extend that technique to the translation of Windows NT
executables into Alpha OpenVMS images.

However, it does require some work in order to get the
application run, and this can't be done by end users.
Moreover, binary translation involves remapping of data which
can reveal hidden bugs.

Activating NT executables on OpenVMS

Given the lots of similarities between WIndows NT and OpenVMS,
it is conceivable to directly activate an NT/Alpha executable
under OpenVMS/Alpha.

    Windows NT address space is very similar to OpenVMS/VAX
 address space which is included in OpenVMS/Alpha address
 space.
    The OpenVMS image activator can be modified to be able to
recognize native Windows NT executable files.
    Windows DLLs are very similar to OpenVMS shareable
 images.
    Standard Win 32 API is already provided as part of the
affinity program.


Being more ambitious: Digital in the next years
-----------------------------------------------

Learn from the recent past

The request for more standards has led to a great consumption
of development resources. As de juro standards are always a
matter of tradeoffs between opposite concerns and commercial
issues, the net result tend to be a real monster (OSI
networking, POSIX, etc...). This never makes the computer
industry progress in terms of innovation. The time devoted to
standard has never helped improve productivity and often has
even reduced it (monsters are not that easy to domesticate).

Years ago, Digital used to be an innovative company. It has
spent the last few years running behind standards. It has
loosed its uniqueness and consequently a lot of its customers.
Not taking that into account will lead Digital to only one
activity : selling boxes. This doesn't require as many people
as Digital now has; so, be prepared for more layoffs.


Invent a new style of computing

With the ability to run on OpenVMS platforms any Windows
application bought directly from the nearest computer shop
just round the corner, a network of X-terminals linked to an
OpenVMS cluster could be a very strong, high availability,
very easy to manage alternative to the hardly manageable, and
support efforts consuming networks of PCs we all know about.
An ethernet cluster of workstations (formerly known as LAVc)
is the world most easy to manage, most secure and most
available configuration. This makes it in fact the most cost
effective solution because it has no hidden costs.

UNIX is now impaired by its lack of personal productivity
applications and its poor user interface.  Almost every
analyst is predicting a decrease in UNIX market share. Windows
NT is lacking a great bunch of functionalities and will
probably need from 5 to 10 years to be a mature and fully
featured operating system. OpenVMS has a role to play because
it can conciliate those two different worlds with such things
as 1170 API, Win 32 API and Internet Ready software.


The key to the success

Digital should immediately stop doing the following :
    taking directions that are only dictated by the will of
 making the predictions of some analysts happen
    staying in the shadow of Mr Bill - Digital must stand
apart from Microsoft unless it wants to end up selling boxes
with less and less employees, narrowing margins and declining
revenues

Digital  should concentrate on delivering more and more base
operating system functionalities because only analysts are
stupid enough to think the operating system is no longer the
problem (Windows 95 is a good example of how operating system
evolution is a key to sales and profit).

Digital should communicate on OpenVMS unique functionalities
and continue to invent the style of computing of tomorrow.
Digital should learn from the Microsoft example that the most
important thing is to make people dream...





	Jean-Pierre PETIT
	DECUS France
	VMS systems SIG chair
	petit_jp@decus.fr



