Ian & Stuart's Australian Mac 1993 September/September 93.iso/Apple/Futures/Kaleida & Taligent/Taligent A New Paradigm

Text File | 1993-10-12 | 29.1 KB | 576 lines | [TEXT/R*ch]

TALIGENT: A NEW PARADIGM, A NEW APPROACH
By Bob Orfali and Dan Harkey

To OS/2 developers, Taligent can seem like a "Land of Oz" of
operating systems, located "somewhere over the rainbow."  Why worry
about? Taligent when our brain cells are overloaded just trying to
digest new OS/2 features like DSOM, the microkernel, DCE, LAN
NetView, and motion-video multimedia?  Why should OS/2 developers
like us need to know about Taligent?  At least that was our attitude
when we began this article--but it soon became evident that there
was a lot about Taligent that OS/2 developers need to know about.

[
I found this paragraph a strange way to start the article... it's
clear that most OS/2 developers outside IBM are very cognizant of
Taligent... did they think this paragraph would fool them into
believing they're the only one interested?
IBM (the OS/2 portion) knows of this interest... as prior to this they
have take every opportunity to downplay Taligent.
]

A NEW PARADIGM?

The first thing we discovered about Taligent is that its technology
is multifaceted, elusive, and would not fit neatly into our
current model of "the way things are."  For this article we were
not provided with specifications, product sheets, demos, or an
architecture reference model.  And this operating system is not
like any other; it's all new and different.  When we expressed our
bewilderment to Mike Potel, Taligent's vice president of technology,
he joked, "Joe Guglielmi had the same problem when he became CEO
at Taligent.  Every time he talked to somebody, he got a different
view of this beast."

This article starts by looking at Taligent from an OS/2 perspective
and provides a framework for the terminology of the interview.  We
then turn the microphone over to Joe Guglielmi.  We could never
match Joe's eloquence and passion for his product.

OS/2 AND WORKPLACE OS

When does Taligent enter the picture?  In Figure 1, we offer a
"fearless forecast" of how the desktop operating systems from IBM
and Taligent come into the picture over the next few years.  In
all cases, the future lies in operating systems that embrace
object-oriented technology.  OS2 2.x is optimized for Intel processors
and will continue to evolve throughout the '90s.  The Workplace
OS is a portable environment that is designed to run on top of
Intel and RISC hardware from a variety of vendors, with the
portability layer provided by a common microkernel.  The 32-bit
OS/2 applications you create should run on all platforms.

In our scenario, objects become very important.  Distributed object
services based on the System Object Model (SOM) become a vehicle
to share objects across operating systems such as OS/2, DOS, Windows,
UNIX, AS/400, and MVS.  In addition, object-oriented application
frameworks containing both IBM and Taligent technology will be
offered for OS/2 2.x, UNIX, and Workplace OS.

In Figures 2 and 3, OS/2 2.x and the Workplace OS support DSOM,
object services, and the object-oriented application frameworks.
OS/2 2.x and the Workplace OS support DSOM, object services, and
the object-oriented application frameworks.  OS/2 2.x continues to
evolve, providing optimal performance on the large installed base
of Intel machines.

A GUIDE TO MICROKERNELS, PERSONALITIES, AND FRAMEWORKS

Figure 3 helps explain all the new terminology and shows where
Taligent fits in.  Several pieces of the Workplace OS play together;
the IBM microkernel, the operating system personalities, the
distributed object layer, the object-oriented frameworks containing
Taligent technology, and the Workplace user interface.  We will
briefly introduce these pieces, then move on to Taligent.

THE IBM MICROKERNEL utilizes technology from Mach 3.0 research.
Mach 3.0 is a portable microkernel developed by Carnegie Mellon
University.  The IBM microkernel and Mach 3.0 remove the
UNIX-specific elements and provide a few well-defined services--including
interprocess communications, virtual memory, ports, task dispatching,
and threads--used to build all system services.  The rest of the
operating system functions--file I/O, user interface services,
device drivers, and communications--are implemented outside the
microkernel.  As a result, the microkernel can be smaller, faster,
and more scalable while providing a higher level of robustness,
security, and integrity.  Elements outside the microkernel, such
as the device drivers and file systems, can be shared by the
operating systems that run on top of it.  IBM's multithreaded
microkernel supports symmetric multiprocessing, a technique that
allows programs to run concurrently on multiple processors in
tightly-coupled configurations.  It is intended to support Intel
and a variety of RISC processors.

TALIGENT AT A GLANCE

Founded in March 1992, Taligent is an independent system software
company owned jointly by Apple Computer Inc. and IBM Corp. and
headquartered in Cupertino, Calif.  The company has 310 employees,
who use object-oriented technology to develop system software from
the bottom up.  The emphasis of the company's charter is on enabling
the innovators and entrepreneurs who spawned the desktop revolution,
allowing them to take full advantage of object technology's benefits
and encouraging a new model for innovation centered around objects.
Taligent's system software will be open for extension at all levels
by software developers, hardware OEMs, and systems vendors.  The
company will license, market, and support its software platform
worldwide.

PERSONALITY MODULES provide operating system-specific API services.
IBM intends to support DOS, Windows 3.x, OS/2 2.x, and UNIX
personalities, with the microkernel concurrently executive multiple
operating system personalities on one machine.  (One personality,
designated as dominant, controls the appearance of the desktop.)
The personalities will preserve investments in code and application
packages.  The microkernel also provides the means to build spec-
ialized servers and operating system replacements.

THE DISTRIBUTED OBJECT LAYER, based on IBM's Distributed System
Object Model (DSOM) technology, allows objects to operate across
networks.  This layer also includes services for storing, replicating,
shadowing, creating, destroying, and specifying objects.  DSOM is
based on standards set by the Object Management Group (ONG), an
industry consortium of over 200 member companies.  For example,
DSOM defines objects using the Common Object Request Broker Architec-
ture (CORBA) interface definition language.  DSOM uses OMG's
specification of an Object Request Broker (ORB) to find and invoke
objects on different machines.  The object services are intended
to comply with OMG's object life cycle and persistence specifications
when they become available.  OMG and IBM are also working on specs
for shared objects and object transactions for client/server
environments.  All this means that OS/2 is already deeply involved
in objects, adhering to current industry standards.  These objects
are, of course, built on top of a standard operating system.

THE OBJECT-ORIENTED APPLICATION FRAMEWORKS intend to provide a
portable distributed set of object services to help create user
applications.  A framework makes writing object-oriented applications
from raw class libraries less tedious; you begin with complete
working subsystems, which can be customized for individual
applications.  You can also create applications by writing subsystems
together with visual application assembly tools.  Third party
software companies will be able to provide parts or entire components
that can be assembled or modified by users or system integrators.

THE USER INTERFACE for IBM's "Workplace" line is (you may have
guessed from the name) based on OS/2's Workplace Shell.  This
object-oriented user interface will be adapted for DOS, UNIX, and
OS/2.  Users think in terms of directly manipulating objects on
the desktop rather than dealing with programs and other computer-
based metaphors.  IBM is working on making the user interface even
easier to use by adding powerful new visual metaphors.

Now that we've introduced some of the terminology, we give the
floor to Joe Guglielmi, the chairman and CEO of Taligent.  We used
the system detailed in Figure 3 as a reference point for our
questions.

_Developer: Let's start with one of the loftier questions. Taligent
was jointly founded by IBM and Apple.  What were your original
goals, and how have they changed?_

Joe: When Apple and IBM got together on this project, it was because
both companies had a common vision of the importance that
object-oriented technology would have in the marketplace. Both had
substantial interest and investments already in that technology.
We first looked at a joint development using the Power PC RISC chip
and then it grew into a more encompassing project that would com-
mercialize objects on high-volume platforms and get the industry
to make the transition from a procedural world to an object world.

_Why is that so important?_

Because our goal is to provide a dramatic improvement in the
application development environment.  I talk about it in terms of
moving the cycle from years to months and providing a brand new
environment, in terms of functionality, that will encourage
innovation. It also turns out that, since we're writing an operating
system from scratch, we get to do some things right.  We now have
a rich heritage in multimedia, advanced graphics support, and other
capabilities that we know should be fully integrated into an
operating form.  So as a kind of a Luck Strike Extra, you get to
do those in an integrated fashion.  By the way, in an object system
all those things become simply objects that you can deal with.
You don't worry about whether you're dealing with full-motion
video or whether you're dealing with static date types.  Everything
is treated as an object structure.

_Where do you see this technology going?_

We see enormous potential in this technology--not because it's
object-oriented, but because it can give the industry a brand new
base that will allow us to move from today's environments, which
are constrained, we think by the operating system capabilities.
Just as OS/2 2.0 unleashed the power of a microchip we'd been
shipping for years, this will begin to unleash the power and
creativity in the industry in a different way.  We won't be
constrained by the old paradigms of procedural operating systems.

_How will you get this technology into the marketplace?_

This is where our strategy has changed dramatically.  Initially,
the notion was to deliver one large brand-new operating system,
top to bottom, in one really significant drop.  We've become more
realistic about that; we've had to deal with very pragmatic issues.
How do we, when providing this great step function, deal with
current investments?  How are we going to make sure the OS/2 invest-
ments, AIX investments, System 7 investments carry forward?  How
do we ensure that we don't force a change in the marketplace all
at once to a new technology?

_So how do you do that?_

We've focused our strategy over the last year on three or four
things.  One is moving the project from a research project with
two or three technology bases to a product development environment
where we now have the focus on delivering products that respond to
customers' requirements to the marketplace.  Second, we've spent
a lot of time staging the project...we can't do it all at once.
We'll never be able, in one release, to compare to those established
operating systems, so that's a bad goal.  So let's focus on areas
in which we think the technology substantially leverages itself.
We're going to stage our technology to focus on both corporate and
software developers.  What do they need first, to provide value?
Taligent's proposition is that the value comes from the devel-
opers.  We don't compete with them for word processors or spreadsheets,
and we don't do databases. What we provide is an environment that
lets them do a much better implementation of what they do best.
Our delivery channels--Apple, IBM, and others--are also going to
add value to this.

_Yes, but every new operating system needs to do that kind of stuff..._

It does.  But the second new notion is that we would take pieces
of this technology and make it available to current operating
systems like OS/2, AIX, and System 7.  That's a pretty profound
change.  That's something that's not very natural for a team
developing a new operating system to do. A development team wants
to make the biggest step function change, one that differentiates
between current and future environments.  Well, we concluded that
without some strategy to get the technology into the market early,
the step function was too great.  This isn't 1983 anymore.  This
isn't about "I have a great technology; we'll just get people to
stop what they're using and move over to it."  We have example
after example in the marketplace today where great technology can't
get commercial acceptance. So the notion was, why don't we take
pieces of this technology and make them available to current
operating systems?

_What does that do for those operating systems?_

First of all, if we've done classy implementations of the technology,
it will enhance them. OS/2 will be more competitive; System 7 will
be more competitive.  By the way, we're not limiting it to just
those two; our strategy is to go after as many of the current
operating systems as possible.  Think about all the UNIX environments
or--ultimately--any 32-bit system that can carry the technology.

_What are the benefits to Taligent of making this technology available to
others?_

The technology will make OS/2 and System 7 more competitive and
they, in turn, will provide us a transition platform for the
wide-scale introduction of objects to the marketplace.  We benefit
if IBM and Apple evangelize object technologies and provide us
with a more evolutionary path.  We're a small company; we can't do
it all ourselves.  But by putting pieces out in the marketplace,
we begin to provide a transition mechanism--that is, investments
made in applications that use that technology under a mature
operating system can be carried forward into Taligent when the time
is right.  So when Taligent comes, customers can decide whether
they want to move to it or not.  this is not a forced march.  It
doesn't mean one Monday morning OS/2 goes out of business or System
7 goes out of business--that's not the strategy.

_But if you've made your object technology available on other
platforms, why would anybody ever want to move to Taligent?_

Taligent will be fully object-oriented with a very consistent object
model.  When you enter Taligent, everything in the system is an
object, with no differentiation between a system object and an
application object.  We can implement a very consistent object
structure once you're in the object space, you can leverage every
element in that environment.  You won't ever have to deal with the
procedural elements of a system.  Take NextStep, for example.
Steve (Jobs) has a great user interface tool, but you still have to
bounce into UNIX from within your applications.

_Where are the advantages of having objects at the operating system level?"_

Envision a growing industry that would provide system frameworks,
networking frameworks.  Take communications.  If they don't like
ours, IBM or Novell could replace it with their own implementation.
Device adapters can inherit characteristics of a particular device
class and customize them to fit their particular needs.  It's a
very consistent way of taking advantage of new hardware.  System
software can then keep up with fast-moving hardware technology.

_OK, so what's in it for application software developers?_

If we do our job right, in the mid-90's application developers will
not be writing code but will instead be shopping for objects.  If
you think about it, there will be thousands of objects; how do I
know which one I want to subclass or reuse?  We're building a very
consistent development environment with viewers or browsers that
let you try out these objects and understand what they can do for
you.  You'll have a full environment that lets you think of your
world in terms of objects.

_Will some of that environment be made available on OS.2?_

Yes.  I didn't mean to imply that the layers of components we
provide for OS/2 won't be great; they'll move objects forward and
condition the mass market for objects.  Taligent has the added
luxury of being able to put the whole thing together from the bottom
to the top.

_Speaking of the bottom, both Taligent and OS/2 will be running on
the same Mach 3 microkernel--is that correct?_

We're on the same microkernel, but IBM is not shipping Mach.
They're shipping a microkernel based on Mach technology.  Mach is
a big, heavyweight thing; IBM and Taligent are creating a much more
lightweight version of Mach and creating personality-neutral servers
and services.  Taligent is adding the mechanisms to make it a
first-class object environment.  We're working on the performance
implications of moving thousands of objects around in a system.
IBM will build the microkernel that we'll use as part of our common
strategy.

[
By here it seems clear to me that IBM has forced a total redesign
of Taligent (Pink had it's own Microkernel).
]

_Mach is an open microkernel, something perceived as a benefit.
Aren't you losing that by going to your own microkernel?_

I'm not commenting on the open aspect.  I'm commenting on Mach's
poor performance for doing objects.  We have over four years of
experience writing microkernels for objects, and we know how to
do one.  Object people don't think you can do objects on the current
version of Mach--it may be OK for doing UNIX-type things.  The
common IBM microkernel solves those technical problems.  On the
openness question, it's up to IBM how they supply the microkernel
back to the industry. I'm trying to answer the guy who says, "What
are you, nuts?  Mach can't run an object system the way it's
currently designed."  The answer is "It can't, but the kernel we're
implementing can."  It's really this notion that the core technology
is optimized object technology and this new common microkernel
works across OS/2, UNIX, and ultimately, hopefully, the industry.
That's where the world's going.

[
So much for IBM's "commitment" to Open Systems!
]

_So is this microkernel one of the first pieces to be delivered?_

It's actually being developed at IBM.  Our operating system
development team is working with the IBM team.

_Are your object frameworks built on SOM and CORBA technologies?_

I think we've come to appreciate what SOM is and what it isn't.
I think there's a greater feeling that SOM's design point and the
problem set that it solves are important, very important.  Whether
or not we can, as a team, adopt all of the goals of SOM in our
design point is the discussion we're having with IBM now.  The
reason is, I think, pretty neat to think about.  In Taligent, we
are dealing with a system optimized for a very large number of
small objects.  That's how we get performance.  We don't move a
few big objects.  We move a lot of very little things around, we
we have a tremendous amount of tasking going on.  We've got the
machine burning and we've optimized for performance.  We got a lot
of value out of that optimization.  SOM is best when the primary
objective is mixing objects from a variety of vendors.  The fact
that the object's interface is the standard defined by the Object
Management Group--the CORBA specification--is also very attractive.

[
Translation? IBM forced SOM down our throats, but we're coping.
Clearly SOM things written for OS/2 currently will not be easily
portable to Taligent.
I'm surprised there is no mention of migration facilities... i.e.
how do PM and Mac ToolBox fit in?
]

_Will you support DSOM then, since it's based on CORBA?_

Of course, DSOM is a very easy case.  Its ORB, which is CORBA
compliant, is one way we deal with objects across networks.  We
know there are objects on this network that are going to be moved
around.

_So you're depending on DSOM to bring CORBA.  Is IBM representing
Taligent's position in the OMG?_

IBM, like Sun or HP, has put some time into defining the CORBA
compliance specs.  In our implementation, we support that spec as
one of the important distributed object models in the industry.
Now take Microsoft; they're doing CAIRO.  I don't know if CAIRO is
CORBA compliant; the likelihood is not. What's going to be our
position on Microsoft's model?  If it's a high-volume model in the
future, we're going to have to find a way to support it as much as
possible.  That doesn't mean we like it...or don't like it.  Our
position is pragmatic.  We have to exist with whatever becomes a
high-volume platform in the future.  We're working with IBM, Apple,
and others to try to make the transition as seamless as we can,
but we're coming from two different worlds.  What we won't do is
suboptimize our world.  If we end up lowering our goal of full
utilization of object technology, we will lose the great differentiation
that comes from it.
[
Note, he does NOT say that they'll support DSOM.
]

_Which IBM groups do you work with on object technology?_

I deal through the PSP division, which means that Larry Loucks,
Cliff Reeves, and Lee Reiswig are my contacts in the IBM world.
So when I deal with Larry and Cliff, we either agree or don't agree,
and I expect them either to bring the rest of the company with them
or tell me that I've got to deal with the situation differently.
By the way, they're doing a great job evangelizing Taligent within
IBM.

_If you were speaking directly to, let's say, leading edge OS/2
application developers who want to do the right thing and position
their product for the future, what kind of strategy or recommendation
would you map out?_

Here's the strategy I think has to be followed.  If you're an OS/2
developer, keep working with OS/2 release 2.1 and beyond.  This
will give you the highest affinity with Taligent.  Why do I say
that?  Because I'm working very hard to make that happen.  Can I
[
So what happens to Mac developers? (see below)
]
tell developers that everything they do in every release will be
cared for seamlessly?  No, I can't tell them that.  But they should
know that we are working very hard with the PSP development group
to put as much of this technology as possible into OS/2 and AIX,
to enhance their competitiveness wherever we can, because we believe
that's good.  We believe a big OS/2 base is in Taligent's interest
and, therefore, we're going to work as hard as we can with IBM to
make it as painless as we can.  We hope it's zero pain.  That's
our goal.  But it probably won't be.

_How do you balance that with Apple's needs?_

The same way.  When we talk to Apple, I have the same strategy.
The decisions we make optimize on where Taligent is going, mindful;
of the fact that helping make Apple successful with System 7 and
IBM with OS/2, AIX, and Workplace OS is good for us, because every
user in those two or three camps is going to be positioned much
better to come to Taligent.
[
Oh, so both will have the "highest affinity"... but there are huge
differences between the two API sets... it is impossible to "integrate"
them.  Is he just talking air?
]

_Will the Taligent user interface look like Mac, Workplace Shell,
or something brand new?_

That's unclear, frankly, and let me tell you why.  I believe that
the opportunity to build the Taligent desktop completely from an
object paradigm gives us degrees of freedom that can move the
desktop beyond what is possible today.  So we've focused on that;
we're trying to figure out where we can exploit this technology
the best.  Let's take one model that could happen.  In a Taligent
world, since everything is an object, applications don't exist the
way they exist today.  You turn your system on and it comes up with
a set of objects that you deal with, and you arrange them in the
workflows.  The notion of a workflow world centered on the user is
much more realistic in a full-object world.  Everything on the
screen is a full object linked from the top to the bottom of the
system.  When we are ready to deliver a full-object model to the
screen, we're going to go back and work with IBM and others on
where they are today.  We'll try to make it as evolutionary as we
can...we'll find ways to see if we can't evolve multiple releases
of the Workplace Shell or the Mac look to where Taligent is
heading.

_What's your planning horizon?_

We're the only ones in town who are worried about 1996.  I don't
mean that in a negative way.  Everybody else is worried about 1993,
'94, and '95.  I have to stay worried about 1996, or '97; how can
this technology, when it's fully exploited, change the playing
field in the industry?  If we miss the goal of providing dramatic
new technology, the industry won't change.  If the industry
doesn't change, none of the interim operating systems are going to
be successful by themselves.  They will not have the kind of
critical mass that you require.  So the design point is to fully
exploit the technology and then look backwards to see how we move
the current base from here to there, and there are several bases.
But the thing people ought to feel good about is that we're thinking
about it.  It's not like we're over here doing our best thing and
hope that someday it all works.  This is the first time there is
a group that is allowed to really focus on where we want to be.
At IBM, you know that there can be terrible constraints on getting
there when you must suboptimize to support the core installed base.
But we don't want to look like we never talk to each other, either,
so we have a new development model here.  It's different than we've
ever tried before, and that's why Taligent is a separate company.

_Will Taligent technology bring together the System 7 and OS/2 worlds?_

As a separate company, I can talk with a lot of authority about
what we're doing, and I can tell you only a little bit about how
we're working with these two companies.  If you want their view of
the world, you've got to go ask IBM and Apple,  I will tell you
this: both Apple and IBM feel strongly about keeping their current
systems viable, and we support that.  There's no model here, under
any circumstance, where all of a sudden on Monday morning OS/2 or
System 7 goes away.  That isn't in the cards, because there's no
viable scenario where this whole thing can make the transition in
one day to a new environment.  So there is going to be, as you
expect, a continuation of a substantial investment in those current
operating systems, and we're going to find a way to make the path
to Taligent very attractive.

_Are you going to sell a shrink-wrapped Taligent through software stores?_

Yes, but that isn't our model.  The whole distribution model is
changing today.  The store distribution model is going away; I
mean, who distributes operating systems today?  The hardware
manufacturers.  So our strategy is to get the distributors to accept
the technology, to provide the critical mass, the target volumes,
so that developers can see it.  Shrink-wrap and user sales will come
after that.  So the whole model is different.  It isn't about
running big ads in magazines, hoping that users will ask developers
to create a great application.  It's about going to developers and
saying, "Look, you're building a great application.  You can keep
doing it.  We're going to give you the tools to do it in a third
of the time and to maintain it more easily by an order of magnitude.
You know, by the way, that when you upgrade it, your cycle will go
from two years to six months.  That's worth a lot by itself.  Oh,
there's another opportunity coming.  Have you thought about
spreadsheet engines, word processor engines, graphics engines?
Have you thought about work flow as an application?  You can really
do it now.  Have you thought about it?"  So we're creating a
different world of applications and when we start this model, all
of these characteristics will allow hardware manufacturers to
differentiate their products.

We want to sell through the high-volume channels.  If we don't do
that, our adoption curve is going to be seven years.  It takes
seven years to get an operating system to the market, if you're
good.  I want to be a success faster than that.  So all of these
strategies are working toward substantially reducing the adoption
curve for the new technology.

_How do you see the operating system pie divided in 1996?_

I'd be presumptuous to give you a target; I will tell you that
existing systems will continue to have a large market share.
Between now and 1996, there's just not enough time to substantially
change the rate flow.  I don't know what the division between NT,
DOS, OS/2, and Windows will be; who knows? That's completely up to
them.  We haven't see NT yet, but they'll have a large share.  I
believe OS/2 will gain critical mass and have a substantial share.
I think Apple will continue to have a big share.  UNIX will probably
still own eight or nine percent of a growing market.  By 1995
there'll be 35 million micros shipped per year, so even 10% of that
is still a big volume.

And then Taligent will begin to show up.  Now I can't tell you
exactly how big our wedge will be, but it will not be zero, because
we intend to get to market by the mid 90s; hopefully earlier.  And
I think that, as Taligent becomes an important environment and we
demonstrate its utility, that wedge will grow more rapidly than
historical growth rates in new operating systems.