Dow Jones & Company, Inc.

InterOffice Memo


To:
Greg Gerdy
From:
Rex Ballard

Date:
April 26, 1993

Subject:
Industry Watch on Operating Systems.



Attached is a brief summary of the various
operating systems and some of the information reported by users.
 In addition to personal experience with OS/2, Windows, MS-DOS,
and UNIX, interviews with Mike Bird, who has experience with Windows-NT
and the other operating systems yielded a more complete and accurate
picture.

A quick summary of the following operating systems will be expanded
in the attachment.

The PC/XT and PC/AT markets.

MS-DOS - The "Tricks" operating system.

Windows 3.0 - The last chance for older systems such as PC/XT
compatibles.

Windows 3.1 - The last chance for the 80286 PC/AT compatibles.

OS/2 1.3 - Obsolete, Unsupported, and IBM's last PC/AT support

32-bit Operating Systems and a whole new game.

Windows-NT -  Too much (minimum RAM, Disk, CPU Speed), too little
(Single User, Short Featured) Can Microsoft compete?

OS/2 - IBM includes everything, including the kitchen sink, can
they support it?

UNIX - Is a 20 year old operating system about to become an overnight
success?

What's new about UNIX?  MACH, OSF/1, Unovel.

Futures - NeXT Step 486, Talegent

Trends and Breakthroughs to Watch For:

Internet - 2 Million strong and doubling every 6 months.

Open Systems - Establishing a good reputation.

User Supported Software - The "Underground" Information
Systems Economy.

Outgrowing Everything - Large Scale Office Automation demands
new solutions.

Wide Area Information Systems - Wide Area Networks spawn Wide
Area Services.

OSI - Will the Specifications ever be specific?  The Stinger.

Safe Bets - Subsets

 POSIX, FIPS, Cross Platform

CC: Bill Cox, Chris Maloney, Mike Pachter, Mike Rauch
Operating Systems of the 1990's and Beyond



The information industry, especially the personal computer industry
can be very unpredictable.  It is an industry where a seemingly
trivial technological breakthrough can make or break the largest
companies in the world.  We have seen IBM, once one of the largest
computer companies in the world, unchallenged by any competitor,
struggle for its life as companies started from almost nothing
consume almost its entire customer base.  Microsoft, begun as
an attempt to sell a junior high school science project has become
one of the largest companies in the world.  Computer Associates
established itself by outsupporting IBM.  Sun has practically
exploded by capturing the midrange market in workstations and
servers by providing commercial "wrappers" and interfaces
to user supported software.

Attempting to make a prediction of what operating systems will
be major players in the future IS market would be like trying
to predict the stock market based solely on past activity.  On
the other hand, there are certain principles that seem to be consistent.
 There are events which can trigger major shifts in the market:

  The AppleSoft Basic in ROM gave it an edge over the "De-facto"
standard S-100 bus in 1977.  The Tandy TRS-80 provided a price
breakthrough that cut into Apple's market.

The IBM/Intel/Microsoft partnership completely displaced the CP/M
systems which were the "De-Facto" standard in 1981.

The IBM attempt to introduce a tightly licensed MicroChannel architecture,
that caused the formation of EISA and VESA bus support.

The AT&T/SUN alliance and announcement that triggered the
formation of OSF and delayed the universal support of UNIX based
systems by 3 years.

Now we have Microsoft, breaking the "Exclusive Trust"
with Intel, IBM, and third party vendors such as Lotus, WordPerfect,
and Informix (Who are now supporting UNIX).  The partnerships
with MIPS and DEC seem more like an effort to derail WindowsNT,
since both MIPS and DEC have hedged their bets with the UNIX market.

What follows are some of the factors that will help to determine
the future directions, within the next three to eight years. 
The question is not whether, but when these factors will play
themselves out.  Pure economic power was able to extend the IBM/370
MVS life span even when its competitors could perform the same
work faster, cheaper, and more effectively.  Today, "Downsizing"
has become a euphemism reducing or eliminating the 370, COBOL,
and SNA systems in favor of "Open" systems which can
be upgraded, replaced, or reassigned with as little effort as
recompiling source code.  All indications are that a similar shake-out
is about to occur in the Personal Computer market.

PC/XT and PC/AT Markets.

In 1980, Microsoft and IBM found it hard to imagine
that anyone would need more than 1 megabyte of memory, multitasking,
or high speed communications.  As a result, the IBM PC was designed
as a single user, single tasking, 1 megabyte system that used
software distributed on floppy disks and had relatively small
disk drives.  The PC/AT eventually included RS-232 asynchronous
communications ports that could communicate at speeds of up to
38.4 Kilobits per second.  The MS-DOS operating system was a cross
between the CP/M operating system developed by Digital Research
and the RT/11 operating system Bill Gates had used to develop
his BASIC programming language (APPLESOFT, MICROSOFT BASIC).


MS-DOS and the "Tricks"

MS-DOS when it was conceived was intended to simulate
CP/M.  Bill Gates also added some features from RT-11.  The idea
was to keep the operating system extremely simple and use loaded
applications to access the more sophisticated features.  There
was no desire for multitasking, background processing, or multiple
user support.

When the IBM PC came out, IBM sponsored a "contest"
that encouraged software developers to write software that took
advantage of all of IBM's new features including graphics, color,
cursor positioning, and the ability to capture almost any interrupt.
  The new software was interesting.  Since the user could freely
access all memory, I/O ports, and Interrupt vectors, the Operating
System became almost secondary to such features as Terminate and
Stay Resident (TSR) programs, Memory Mapped Video applications,
and copy protection schemes that included software timing loops,
deliberate corruption of disk sectors, and mandatory location
of data on absolute sectors of certain disks.

Because of all these "Tricks", upgrades
in hardware became limited by the software base.  Each upgrade
would be followed by many upset customers.  The PC/XT didn't have
the ROM BASIC to jump into, so software that depended on that
wouldn't work.  The PC/AT ran so quickly that the software timing
loops were corrupted and destroyed hard disk data.  Even though
the PC/AT was capable of addressing 16 Megabytes, programs that
depended on being able to map segments based on 16 byte pages
could not use the enhanced mode.  A bug in the "virtual"
mode was fixed in later releases of the 80286 chip, but not in
time to encourage development of virtual mode applications.  The
"Protected Mode" was used by applications to keep the
Operating System from taking control.  MS-DOS became a set of
multitasking applications started by a monotasking operating system.





Windows 3.0 -- The last chance for older systems such as PC/XT
compatibles.

With Windows 3.0, Microsoft took back control of
the operating system.  Applications that wrote directly to the
screen, captured interrupts, and launched TSR programs were considered
"Ill Behaved", and became undesirable because they could
only be run in a "crippled" mode.  Software developers
scrambled to conform to the new rules.  In the process of porting
to this new standard, they rewrote most of the new software in
portable languages such as C, using portable routines and library
functions.  Within a few months of releasing a package for Windows,
developers would announce a package for OS/2 and a package for
UNIX.  The UNIX releases were usually geared to 80386 and SPARC
(SUN) processors and libraries.  Today, Microsoft and Borland
are conspicuously missing from these markets.  The other operating
systems did not support the 8086 processors.  The OS/2 2.0 release
didn't even support the 80286, and both OS/2 and UNIX required
almost 4 Megabytes to use effectively.  For PC/XT users or PC/AT
users with less that 2 megabytes of RAM, Windows 3.0 was the only
choice available.


OS/2 1.3 -- Obsolete, Unsupported, and IBM's last PC/AT support


OS/2 Version 1 was developed by Microsoft and IBM
to replace the MS-DOS operating system.  The minimum requirements
were an 80286, 4 megabytes of RAM, and very well behaved applications.
 Support from third party vendors was slow in coming, especially
from communications driver vendors and network vendors.  The operating
system support for these features was announced as only being
available in the "Extended Edition" sold exclusively
through IBM sources.  Extended Edition also included the "LAN
Manager", which was direct competition to Novell, Banyan,
and the TCP/IP vendors.  Most of the industry ended up with a
"wait and see" attitude.  When Microsoft walked away
from the joint venture and released Windows 3.0, the support for
OS/2 dwindled.


Windows 3.1 -- The last chance for the 80286 PC/AT compatibles.


With Windows 3.1, Microsoft no longer supported PC/XT
and 8086 machines.  The minimum requirements were up to a 80286
PC/AT or better with at least 2 Megabytes of RAM.  Even CGA support
was inconsistent and required disks from the 3.0 release to support
it.  The recommended driver was the VGA card configured in "Super
VGA" 1024x768 mode.  Windows 3.1 quickly began to look more
like a workstation.


32-bit Operating Systems and a whole new game.


OS/2 - IBM includes everything, including the kitchen sink,
can they support it?

Although OS/2 2.0 was released before Windows 3.1,
it required a minimum of an 80386 processor because it ran most
of the operating system in 32-bit "virtual" mode.  The
graphics were executed in the 16 bit 80286 "Thunk" mode.
 The recommended minimum memory was 4 Megabytes, but an additional
4 megabytes used as disk cache would reduce the time to start
up a new application from about 10 seconds to about 2 seconds.
 The hope was that it would encourage people to trade in their
old 80286 and 8086 machines for PS/2 systems.  IBM supported every
possible software development channel, including Windows 3.0 compatibility.
 Bundled applications included terminal software, spreadsheet,
calendar, E-Mail, and word processing.  The full distribution
takes about 50 megabytes of disk space.  If LAN support is available,
the distribution drops to about 20 megabytes.

The main advantage of OS/2 2.0 is that it was compatible
with almost everything. There were DLLs to support communications,
interprocess communications, LANs using many protocols and drivers.
 They even provided upgrades to user supported software pools
such as the GNU EMACS and GNU compilers.  The EMACS icon is a
"kitchen sink".  This also created the availability
of a full POSIX interface, UNIX applications and APIs.  The only
requirement is access to an Internet host.  Many OS/2 products
are now being shipped over the Internet.  There is even support
for Windows 3.0 applications.

Unfortunately, OS/2 2.0 was released during one of
IBM's weakest quarters.  The sales staff was not structured for
OS/2 standalone sales.  As a result, OS/2 has been fighting several
perceptions and rumors of PS/2 and MicroChannel support that is
not available to ISA and VESA machines (untrue).  The big question
is whether IBM is committed to supporting a mass distribution
of OS/2 to a wide variety of machines.  The user support seems
to be quite broad.  The UNIX "user supported" distributions
work, the DOS Share-ware has been ported, and most Windows 3.0
applications will run without changes.  OS/2 may be a sleeper,
users have been holding out for Windows-NT.


Windows-NT -  Too much (minimum RAM, Disk, CPU Speed), too
little (Single User, Short Featured) Can Microsoft compete?

Windows-NT appears to be a major disappointment.
 The hardware required includes a mandatory 80386 processor, 16
Megabytes of RAM, 70 Megabytes of disk.  It does support multitasking
and even include a skeleton implementation of TCP/IP for use with
UNIX NFS servers.  The TCP/IP package also supports the WinSock
Socket API DLL.  The actual protocol is done by Microsoft.  The
utility functions such as host name to internet address conversion
is done by third party vendors.  This was actually a good compromise
since it gives the flexibility to do NDIS, BIND, X.500, or local
lookup.  Unfortunately, Windows-NT is so large and expensive that
it has sized itself directly into the Workstation, OS/2, and UNIX/Motif
Marketplace.


UNIX - Is a 20 year old operating system about to become an
overnight success?

One of the hottest contenders is UNIX.  It gives
both diversity and compatibility.  There are about 4 "flavors"
of UNIX, and most applications are written to run on all of them.
 The exception software is usually that written to show of the
apparent superiority of their particular flavor.  Applications
developers seeking access to the widest possible marketplace use
compatibility guidelines such as POSIX, ICCCM, FIPS, and DDN/RFC.
 Most of these standards are not declared as standards until they
are available on every type of UNIX.

The UNIX operating system has been around for many
years.  Developed in 1968 as a test and development system for
communications systems, UNIX wasn't even supposed to get out of
the lab.  If the developers had known about Multics (an MIT developed
operating system) they wouldn't have written UNIX at all.  UNIX
has evolved from its early days on a PDP-8 to a PDP/11 (Version
6).  AT&T wasn't able to market UNIX and gave the source code
to some universities, including Massachusetts Institute of Technology
(MIT) and the University of California at Berkeley.

The students at these universities quickly learned
the system and developed simple applications including full screen
editors, filters, several programming languages, and several libraries.
 The interprocess communications were refined by Berkeley (Sockets).
  MIT and Berkeley developed most of the porting tools required
to migrate it from its PDP/11 processor to the VAX, the 68000,
the National NC32000, and even the 8086.  Berkeley added features
such as demand paged virtual memory, shared "Text" and
"Data" areas, the "csh" shells.  MIT created
the Free Software Foundation as a clearing house for user supported
software.  UNIX virtually sold itself.

In 1983, Judge Greene ordered the divestiture of
AT&T and granted them permission to market UNIX as an operating
system.  AT&T could reclaim UNIX completely, and instead traded
a few trade secrets with the University of Berkeley.  The resulting
Berkeley 4.3 distribution was restricted under AT&T agreements
and the AT&T System V.4 release contained all of the Berkeley
4.2 features.  AT&T used this to keep a price floor on both
systems and to limit the amount of kernel development done by
schools.

Around 1987, AT&T declared that the only "official"
ports of UNIX would be to the VAX, the 3Bxx, the 680x0, and the
SPARC processors.  Although this was quickly cleared up, it caused
enough concern that IBM, Hewlett Packard, DEC, Intel, and several
other hardware and software manufacturers formed the Open Software
Foundation in order to prevent AT&T from having a monopoly
on UNIX.  Carnagie-Mellon University, MIT, and several other organizations
had been working on "Freeware" versions of UNIX work-alikes.
 This effort was coordinated and resulted in the development of
the MACH operating system.  Some proprietary enhancements were
added and the result was the OSF/1 operating system.


What's new about UNIX?

The MACH operating system was based on a very simple
kernel that performed scheduling of threads, context switching,
memory allocation and paging of programs, buffers, and shared
libraries.  The UNIX-like operating system, interprocess communications,
and all of the application programmer interfaces were implemented
as shared libraries.  Even the driver software was just a task
that requested special memory.  Interrupt handlers just posted
semaphores.  MACH was small, simple, easy to support and enhance,
and inexpensive.

MACH has become a "baseline" for several
operating systems, including OSF/1, UNIX, Berkeley UNIX (BSD 4.2.1),
and POSIX MACH, but it has also shown itself to be capable of
supporting OS/2, AIX, Windows, and even "Virtual MS-DOS"
BIOS emulation.

In 1991, MIT released the X11/R4 release.  Shortly
after that, the federal government formally adopted the X11/R3
release as a Federal Information Processing Standard (FIPS), which
was supported by X11/R4.  That meant UNIX/POSIX/MACH/OSF had an
official graphics interface.  Up until then, each vendor had its
own graphics API and services.  The "Look and Feel"
were different between various clients, but they would run and
could be displayed on X11 servers.  There are licensing restrictions
on the OSF/MOTIF library and the UI/OpenLook library but they
can coexist, be run as clients, and communicate with each other.
 The portable version (Athena) and contributed versions (HP Widgets)
are often chosen because their simplicity makes them faster on
slow speed servers and links.  The X11/R5 release included 3-D
graphics.  The X11/R6 release officially blessed the InterViews
C++ tool kit which has been available through user supported software
sources since X11/R4.

In effect, the UNIX system has been growing smaller
and smaller, getting faster and faster, and supporting more vendors
and software.  In 1984 UNIX was ported to a PC/XT with 640K RAM,
5 Megabyte disk drive, and Monochrome Graphics Card.  In 1992,
Coherent 3.0 could run on a PC/XT with 640K RAM, 10 Megabytes
Disk Drive (Man Pages included), and a Monochrome Graphics Card,
and cost under $100.  A full featured implementation of UNIX (AIX
2.0 or Interactive) with Motif graphics, Windows 3.0 Compatibility,
Communications Software, NFS Server Support, and full documentation
will run on a 386SX/16 with 8 Megabytes of RAM, 40 Megabytes of
Disk, and a VGA or 8514 card for about $1200.  A shareware version
(Linux) is available for $60.

Novell has announced that it will release a full-featured
implementation of UNIX with X11, MS-DOS emulation, and "Shrink
Wrap Compatibility" (COFF), after April 15 when takes possession
of UNIX from UNIX Systems Laboratories (AT&T).

Mark Williams released a COFF compatible version
of Coherent for $99.  This operating system includes a base development
package, can run shrink-wrapped software written for SCO-UNIX,
SunSoft UNIX, Interactive UNIX, Novel's UNIX, and AT&T System
5.4 UNIX.  It is to the UNIX world what DOS is to the MS-Windows
world.  The windows package is an extra $200, the networking is
an extra $100, the DOS emulation is an extra $100, and so on.
 Coherent does include enough of a "development package"
to port user supported software, and several commercial vendors
are already scrambling to beat the Mark Williams prices or features
(which is relatively easy).

There are several "free" versions of UNIX.
 By free, this means that distributors are not required to pay
royalties, license fees, or franchise fees.  This software is
very robust, often rivaling products from SUN or any other major
vendor.  The software may be downloaded from archives on the internet,
purchased from a "packager", purchased through a "support
consultant", or purchased through retail channels.  There
is little advertising and minimal recording of "sales".
 This approach is very popular when several similar systems are
needed since one effort yields multiple output fields.  User upgrades,
enhancements, and fixes are available through the internet or
through the distributors.  Often, individuals and companies will
"pay" for their software by submitting their user created
software through the internet.  As a result, the software base
is substantial.  Available applications range from CAD/CASE packages
to MS-DOS simulators.  They aren't as "pretty" as the
commercial versions but often they are more useful, especially
when source is available and a few lines of enhancement code make
it a new product.

Three of these versions are worth noting here.  The
BSD386 version is Berkeley 4.3 compatible and supports TCP/IP,
sockets, X11 Windows, and several drivers.  There is no source
code and there are licensing restrictions on some of the applications.
 Linux combines features of BSD, AT&T System V, and Mach.
 It comes in binary or source code format.  Distribution is restricted
only in that distribution may not be restricted.  Anyone who has
Linux can give it to anyone else as long as they let the person
they are giving it to know that it can be given away.  Minix,
the miniature version of UNIX was developed as a teaching aid
is available for 8086 machines and up.  The code is structured
so that it takes advantage of 80286 and 80386 extended addressing
modes.  Minix is often used where text or communications support
is needed and graphics is not.  Minix does not support X11 Windows
and supports very limited graphics and addressable text.
Futures: NeXTStep 486 and Talegent

There is work being done on the NeXT Step software
for the 486.  Current speculation is that it will need more RAM
and disk space than the Windows NT product.  Like the Macintosh,
it comes with a wide variety of bundled software.  The current
pricing structure is prohibitive.  The release date is unknown.
 One of the main advantages was that NextSTEP was an object oriented
version of UNIX.  In 1991, the InterViews toolkit made object
oriented development (C++) available to UNIX.  It has been officially
incorporated into the X11R6 release.



Talegent is the code name for the Joint Venture product
being developed by Apple and IBM.  It is hoped that Talegent will
do for the RISC market what the IBM PC did for the Intel market.
 Details are unknown.
Trends and Breakthroughs to Watch For:

Unfortunately, technical merits often have little
to do with the apparent and actual success or failure of a product.
 Often, a company appears successful based on its ability to advertise
heavily in the right publications and may be cutting staff to
survive (IBM), while other products are virtually unknown in terms
of their television advertising and yet outpace all competitors
 (TCP/IP).  There are some recurring patterns that can be factors
in the overall environment.
Internet -- Two Million strong and doubling every six months.


The Internet was originally part of a restricted
group of government contractors, military agencies, and research
institutions.  In order to qualify for an internet address, you
had to promise to provide useful information to researchers, not
advertise, and to only go into systems that are intended for public
access.  Gradually, students, then graduates, then businesses
got connected and began discovering that the internet was a good
way to communicate.

In 1990, the Bush administration established a framework
for commercializing the Internet.  The goal was simple.  By December
1992, the Internet would be self supporting.  Al Gore announced
a plan to expand the network to almost anyone who wanted access.
 Today, the Internet is doubling in size every 6 months, with
2 million Hosts as of  March 1, 1993.  The NNTP news service,
America's version of a public forum, consists of about 60 megabytes/day
of everything from the latest product announcements to political
opinions from right and left wing extremists.

In 1986, Federal Express popularized the FAX machine
by providing station to station and business to business service.
 The Zap-Mail service actually failed because it was too successful,
the X.25 network could not support the traffic being sent.  The
current estimate is that the TCP/IP version of internet could
support about 60 Million subscribers with current technology.
 There are actually plans for supporting over 2 billion sessions
if necessary (TCP/IP supports multiple sessions/user).  The Internet
is now used for LAN to LAN connection, E-Mail, FAX and Graphics
transfer, software distribution, customer support, and various
public announcements.  It is also read by several government officials
as a "public opinion poll".  Internet is to the computer
what the paved road was to the automobile.  This will have an
impact on not only the computer industry, but also on business
in general.

The interesting thing is that much of this growth
is through restricted distribution such as PSILink at $29-$39
/month which provides Mail, News, and File Transfer, but not interactive
use available to the higher cost $200/month service.  Commercial/Business
users are now replacing their FAX machines with Internet Links.
 The best part is that bandwidth and connectivity are no longer
constraints.  A user can dial-in for as little as $10/month and
a business can connect at up to T1 Speeds for as little as $300/month.
Open Systems - Establishing a good reputation.

In 1982, the Navy conducted a test of interoperability
to determine whether TCP/IP could be used to connect several different
systems ranging from VAX 11/780s to 3090s to PCs to UNIX to Cray.
 The first test was conducted in secret.  The second one became
a public show.  Open system has different meanings to different
people and different organizations.  At minimum it assures that
data can be transferred in a useful manner from one system to
another regardless of system type, architecture, or model.  To
the Free Software Foundation, it means that any application that
can generate data should be available on any machine the data
owner desires without additional charge.  The concept and the
practice of Open Systems architecture have become a set of "de
facto" standards, formal standards, and informal standards.
 Generally, these standards are available to anyone who wants
them with sufficient detail for implementation for little or no
money.  In many cases, "reference" copies are made available
through U.S. government archives.
User Supported Software -- The "Underground" Information
Systems Economy.

Many corporations avoid user supported software because
it was once software created by hobbyists and students.  In many
cases, the quality was poor.  There was also a concern that viruses
and worms would be created and would corrupt the system.  Today
user supported software is software that is submitted by one company,
organization, or individual and is ported, enhanced and supported
by other companies, organizations, or individuals.  Sun is a company
built almost entirely on user supported software.  They are also
a major contributor of user supported software.

There are three ways to handle user supported software.
 You can buy a ported, compiled, built, and self installing version
for $300/workstation.  You can have a contractor unload a tape,
install the software, and leave the source for your later use
for about $1000/product.  Finally, you can have an interested
employee download the source from an archive,  compile and install
it in his spare time, and have an operator install it on a common
server when it has been verified.  For only one client, it might
be better to use the shrink wrap products.  For over 100 users,
it's worth paying the enterprising young man a bonus.  100 copies
of a $400 application is $40,000.  Upgrades are distributed or
announced over the internet.  The biggest advantage is that the
source can be used to port to POSIX or UNIX compatible systems
ranging from a 386SX/16 to Cray 3 and Convex.  The competition
in this market is fierce -- the consumer wins.

The internet news service is bi-directional and often
provides useful insights and quick answers, often from other users.
 Many times, questions are answered publicly.  It looks bad when
a vendor apparently ignores a request for assistance.  Many companies
actually have customer service personnel working with news posters
exclusively.  Many companies discourage the use of telephone support
in preference to this type of service.  In addition, several of
the commercial boards such as Compuserve, Delphi, GEnie, and Dow
Jones are now available over internet.  Common carriers including
Sprint, MCI, and IBM are replacing PADs with Internet compatible
switches.  These switches actually provide bridging between Async,
X.25, CLNS Frame-Relay (OSI), and TCP/IP. 
Outgrowing Everything - Large Scale Office Automation demands
new solutions.

There is an interesting phenomenon that occurs when
companies migrate through the different stages of office automation.
 Initially, terminals connected to a mainframe are replaced by
personal computers.  Later, the PCs are connected with Ethernet
into a simple NetBIOS LAN.  Later, the system outgrows the single
strand and a Network Operating system such as NetWare or 3-Com
XNS is installed.  Soon, the organization grows, electronic mail
becomes a standard practice, and networks are connected with TCP/IP.
 Soon, the user wants to access services instead of just files
and finds themselves using servers and peers.  Eventually, uucp
mail, FAX, and modems are placed on the LAN.   The ability to
pass complex documents, quickly pick through large files, and
access the right data becomes an essential part of the operation
of the business.  This is a good example of why REACH customers
are such heavy users.
Wide Area Information Systems - Wide Area Networks spawn Wide
Area Services.

Eventually, it becomes desirable to get real time
responses to requests for information.  Who has what information?
 Who has what product at what price?  Networks connected to the
internet are often provided with the ability to answer questions
like these.  Instead of every vendor sending his prices to a central
database, he just notifies an "agent" that he would
like to receive these requests.  Requests sent to that agent are
propagated to all of the subscribing "servers".  The
servers send responses directly back to the clients.  There are
several such services that are defined as part of the Mulitcast
specification for Internet.  One company, WAIS is actually developing
"Metacomputing" systems that enable customers to serve
and collect information ranging from price quotes to searches
for a user name.  These systems were originally intended to simplify
name binding services but were abstracted into general purpose
databases.

This type of service will quickly expand to EDI transactions,
including electronic funds transfer, and NASDAQ style trading
of everything from Stocks and Bonds to Computers, Cars, Houses,
Dates, and Businesses.  The interesting thing is that TCP/IP and
Internet were designed to be both wide open and extremely secure.
 When a service requires security, it can be as secure as a bank
vault, and still appear to be easy to use.
OSI - Will the Specifications ever be specific?  The Stinger.


There have been several attempts to create Open Systems
standards committees.   These would include ISO, OSF, and UNIX
International (UI).  There is a paradox here.  In order to become
a standard, the standard must be specific enough to implement
a reference model from the specification.  If the standard is
that specific, there is no competitive advantage to adopting the
standard.  If the standard is not adopted however, the competition
will adopt another standard and will combine as one effective
unit with which you cannot compete.  If you try to license or
excessively restrict the standard, no one will adopt it, or they
will adopt another standard.  If your standard is adopted, you
can attempt to sue, but if you win the suit, the competition will
drop your standard.

This means good news and bad news.  The bad news
is that prices go down due to the intense competition.  The standards
in the UNIX market have driven the UNIX market to where systems
of 1990 are about 1000 times more performance/price than 1986.
 The proprietary markets (VMS, MVS, MSDOS) have seen much smaller
increases and a rapid loss of market share.  The good news is
that the cost of production goes down significantly.  A system
that would have taken 100 man/years to develop using proprietary
operating systems and communications protocols can be produced
in 2 or 3 man months using "off the shelf" reference
code and products. 


Survival

As mentioned, UNIX and Open Systems have a reputation
for selling themselves.  Many times, the industry will rally around
UNIX as a way to compete with a much larger competitor.  UNIX
helped turn the VAX into a system that eventually eliminated the
IBM Minicomputer.  UNIX helped Sun capture the System 3X market.
 UNIX helped Cray and Hitachi capture the Mainframe and Super
Computer markets.  UNIX provided an immediate software base for
the RISC processors that produced breakthrough speeds.  Many vendors
view Microsoft as a threat to their businesses.  For many, they
see UNIX the only alternative to the Windows-NT "Bundles"
that could put them out of business.

UNIX is also a very hard system to compete with.
 IBM tried to compete in 1982 with its Series-1 Mini and eventually
shut down that business.  DEC tried to promote VMS and even when
it was included with the hardware, buyers routinely put UNIX in
its place.  IBM has been cutting down their mainframe MVS and
VM units because they can't compete against the UNIX markets.
  Many customers have refused to upgrade.  Others have stopped
buying the vendor support.  In each case, the solution has been
to respond with a version of UNIX.  DEC released ULTRIX.  IBM
released AIX.  Apple even released AUX.

The most interesting phenomenon is when companies
have released enhanced open systems.  When SUN put graphics on
UNIX, they instantly became a major player.  When SCO and Interactive
supported shrink wrapped software, they opened up a market as
NFS servers and communications servers.  Very few of those who
were "aggressively compatible" suffered.

This has also worked for software vendors.  Lotus
1-2-3, Word Perfect, Oracle and Informix are examples of products
succesfully migrated to open systems with enhanced features. 
Many UNIX software specialists, such as Asterix, CACI, and Island
Graphics have thrived exclusively on the open market.
Safe Bets

In the midst of all this variety, there are actually
some safe bets.  Windows NT, OS/2, UNIX, and AUX are now all POSIX
compliant.  In addition there are cross-platform development tools,
such as XVT, that take the same input and generate code for Windows,
OS/2, Motif, and OpenLook.  Hardware and OS vendors often tout
features that differentiate their product as a way of increasing
the profit margin.  For the application developer, this can be
a trap.   Lotus, once notorious for directly manipulating the
video, has now discovered that they can reach a larger market
in less time and at lower cost by using an appropriate subset
of Open System functions.


Is WindowsNT the future?

There is something interesting worth noting about
all of the alliances related to WindowsNT.  Microsoft is the only
company commited exclusively to NT.  DEC, MIPS, and Intel are
all major players in the UNIX market.  If  NT captures the majority
of the current windows market, they will try to get higher profit
margins by raising the prices.  Unix will quickly become the preferred
choice.  Intel has been enjoying healthy profit margins by having
a monopoly on the 80486 market.

If Microsoft is willing to commit 100% of it's resources
to the success of WindowsNT, including taking very small profit
margins, providing a great deal of advertizing co-op for exclusion
of  mention of UNIX products in 3rd party advertizing, and can
maintain the appearance of being a "De-Facto Standard",
they can keep the press coverage focust on them long enough to
either (1) Come out with their own version of UNIX, (2)  Announce
an even "better" version of Windows that will stall
the UNIX market for another 6 to 9 months.