Date : Sat, 06 Sep 1986 13:39:33-MDT
From : Rick Conn <RCONN@SIMTEL20.ARPA>
Subject: Some data on ZCPR 3.3
The New Z System
by Richard Conn, 6 Sep 1986
I have been working quietly on the development of the new Z System
for quite some time, and I feel that I must speak out now on what I am doing
so that people will know the true story. This presentation is short and not
complete, but it is enough to answer some basic questions without taking too
much of my time away from the development activities.
Why create a new Z System?
First of all, why am I creating a new Z System? There are many
reasons:
1) I have studied the old Z System for over two years now,
reviewing it for flaws and weaknesses. From the beginning, I have stated
that the Z System is evolutionary, and I have enough new ideas now to implement
the next step in this evolution.
2) I have been learning many new things about other
environments, including Ada Programming Support Environments (APSEs),
UNIX, SUN window-based systems, TOPS-20, VAX/VMS, and Symbolics workstations.
I have seen many good ideas that I would like to incorporate into my
personal computer environment, the Z System.
3) My next real goal is the development of a banked Z System
followed by a multitasking Z System. An improvement in the current non-banked
Z System is a logical step along the way, giving me a platform from which to
experiment with some of the new ideas before fully incorporating them into
the new systems. ZCPR 3.3 is not a goal, but is a milestone along the road
to a goal.
4) My needs have changed and are continuing to change, and
the old Z System no longer meets my needs. I have to change the system in
order to continue advancing toward my other goals.
At a recent talk I gave at the Trenton Computer Festival, someone
asked me if I plan to move over to the IBM community. If I was motivated
by profit to a large extent, my answer would obviously have been yes. But
my principal motivation is to learn and grow and to have fun doing it. The
money is a very nice side effect, but it is not the driving side effect.
I definitely intend to keep learning through the Z System. The Z System,
coupled with the very rich computer environment I am accessing through
my communications system, has enough potential to meet my needs for many
years to come.
New Standards
Good software engineering principles and common sense indicate that
a lot of benefit can be derived through standards. The evolving ZCPR 3.3
and Z System now have a number of standards going for them:
1) a common language, implemented by the ZAS assembler
2) a common support library, implemented by SYSLIB, Z3LIB,
and VLIB
3) a common format for all assembly language programs,
implemented by PPAL - the Pretty Printer for
Assembly Language
4) a common operating system and support base, implemented
by ZCPR3 and ZRDOS
5) a common documentation format which allows easier
upgrading of the documentation as the programs
change; this system is for both hardcopy (MAN)
and online (HELP) mediums
6) a common configuration management system, CONFIG,
which allows users to easily identify their
configuration and determine if all of their
programs are current
7) page-relocatable structures for system segments,
ZCPR 3.3 itself, and ZRDOS, which server to
further support transportability and make
installation easier
8) a very robust communications protocol, supporting
packet sizes from 1 data byte to 4K data bytes
You will see all of these in the months to come.
The Libraries
One key point to make here is that I purposely did not mention
a common binary structure, like the Environment Descriptor. The Environment
Descriptor is not a standard, but the interface to it, namely Z3LIB and VLIB,
is. This was purposely done so that future growth would be supported.
As the Z System evolves, it will probably be necessary to delete old features
which are no longer needed in favor of new features which provide new
capabilities. If the binary structures, like the Environment Descriptor,
are to preserved, then the growth will be hampered by the lack of availability
of space in which to implement the new features. By standardizing on the
interface to the Environment Descriptor and not the structure of the
descriptor itself, then the Environment Descriptor can take any desired form
without concern for the impact on other programs. Specifically,
ld hl,envadr ; get base address of env
ld de,offflag ; get offset to a flag
add hl,de ; point to the flag
ld a,(hl) ; get the flag value
would be severly impacted if the structure of the Environment Descriptor
changed, but a call like:
ext getflag ; reference library routine to get flag
...
call getflag ; return flag in A
would not be impacted at all by such a change. The library containing the
GETFLAG routine would be impacted, but it is a lot easier to change one
module in a library and reassemble 100 programs than it is to track the
necessary changes in, edit, and then reassemble 100 programs.
With one exception, the routines in the new Z3LIB for ZCPR 3.3
have the same names and interfaces as the routines in the old Z3LIB.
Conversion is simple - reassembly is all that is required.
Unfortunately, many programmers have not yet learned this lesson.
They continue to write code which does not use the libraries. Imagine
the work that will be required to go through their programs, modifying the
references in order to be compatable with the banked ZCPR3. Imagine them
going through the work again for the multitasking ZCPR3, and again and again
as ZCPR3 continues to evolve. Thanks to Echelon, ZCPR 3.3 is still
compatable at the binary level in terms of the Environment Descriptor.
Certain values have been replaced with others, so changes will still be
necessary, but many of the key values have stayed in the same places with
the same meanings. My original approach was to free up all of the dead
space left by the removed values, making more room at the end for growth,
but Echelon convinced me that the impact on programs written by those who
didn't understand the basic principle of the libraries would be devastating.
Thanks to the new design, it only took me 3 hours to convert back with no
loss of functionality -- yet another benefit from the new structured design
of the system -- but I can guarantee that the future systems will not be
converted in this manner.
If everyone starts thinking in terms of the libraries, the future
growth of the Z System is assured. If people cling to the old ideas of not
using the libraries, then it is clear that growth will be curtailed.
The Bottom Line
I am moving forward with the development of the new ZCPR 3.3 and
the new Z System. As you can see, there are many changes in it, and I have
barely skimmed the surface. I will move ahead and release it only when I am
satisfied that it is usable to others. Documentation must go with the
release, so the release will not take place until the documentation is ready.
My goal is to meet my needs, as outlined above, and this goal will be met.
I thank those of you who have been supportive in the efforts of Echelon and
myself, and I sincerely hope that you will share in my enthusiasm for the
new ZCPR 3.3 when I release it to you.
Rick Conn