W8SDZ@SIMTEL20.ARPA (Keith Petersen) (09/16/86)
I just received the following file on my RCP/M.
--Keith Petersen
Arpa: W8SDZ@SIMTEL20.ARPA
uucp: {ihnp4,allegra,cmcl2,decvax,mcnc,mcvax,vax135}!seismo!SIMTEL20.ARPA!w8sdz
GEnie Mail: W8SDZ
RCP/M Royal Oak: 313-759-6569
--cut here--ZAS-SLR2.DOC--cut here--
THE LIBRARIES, ZAS, AND THE SLR ASSEMBLERS -- PART 2
Jay Sage
September 11, 1986
Background
After reading Richard Conn's article on the libraries in Micro/Systems
Journal (MSJ), I felt that Richard had been highly misleading in his
statements such as the following:
SYSLIB Version 3.6 (the most recent version) is written in the
ZAS assembler language and can only be assembled by the ZAS
assembler (SYSLIB.REL is provided in the distribution, so it is
not necessary to be able to assemble SYSLIB in order to use it).
This statement and an identical one about Z3LIB Version 1.3 imply that
anyone who wants to work with the library source code has to go out and buy
ZAS. I felt that this impression needed to be corrected, especially where
(in my opinion) ZAS is at best a mediocre assembler, particularly in
contrast to the superb assemblers from SLR systems that are available at
nearly the same price. I released the file ZAS-SLR.DOC to clarify matters.
Richard has now released his rebuttal in ZAS-STD.DOC. While Richard's
comments in MSJ were merely misleading, some of those in ZAS-STD.DOC could
well be construed as an affront to what is left of the 8-bit programmer
community. More on that a little later. First I would like to deal
specifically with Richard's comments in rebuttal to mine.
Rebuttal to Richard's Rebuttal
In Richard's responses to my comments I thought I was reading something
composed by the State Department -- a lot of semantic sidestepping to
distract one's attention from the real issue, the misleading statements in
the MSJ article.
First Richard corrects me concerning his employment by Echelon. OK,
let's use the words "very closely associated, including financially"
instead. Actually, I may indeed have been wrong. I had been told several
months ago by someone definitely in a position to know that Richard was
working full time for Echelon. Perhaps I misunderstood; perhaps things did
not work out as planned; perhaps the legal relationship is technically not
employment. It really doesn't matter. The real issue is that Richard is
using ZAS not because he examined a lot of assemblers and found ZAS to be
the best. He uses it, I submit, because it is the product promoted by
Echelon. AND THERE IS NOTHING WRONG WITH THAT. There would also be nothing
wrong with his being employed by Echelon and supporting his employer by
using and promoting the employer's products (the fact that Richard makes no
money directly from the sales of ZAS is irrelevant; I don't make any money
directly from my employer's sales either). What is wrong is to promote
those products in a misleading way.
Richard pointed gleefully to my admission that SLR's Z80ASM assembler
could not fully assemble the SYSLIB source as released. I ask the rest of
the community out there: Do you think that, in the nearly three hundred
modules and thousands of lines of code, five lines with ".IN LUDDEF" that
have to be changed to "INCLUDE LUDDEF.LIB" justify conveying the impression
that you better go out and buy ZAS if you want to use the source code? I
find it very hard to believe that anyone who would know what to do with the
source code would have any trouble whatsoever converting the source back to
one of the standard assembler formats.
Richard states that SLR's Z80ASM cannot assemble the latest Z3LIB
without completely editing the source (though in the same text he also
states that he has never seen or used the SLR assembler!). In fact, Z80ASM
assembles the latest releases of Z3LIB (1.3) and VLIB (1.1) without any
problems at all. Perhaps Richard was referring to a future version of the
libraries. Is he planning to make deliberate use of idiosyncrasies of the
ZAS assembler to make sure that only ZAS can be used? Even then I am sure
there would be no serious difficulty in converting the code to a format
suitable for conventional assemblers.
To sum all this up, I repeat my previous statement that if Richard had
said something like "the most recent versions of the libraries are written
in the ZAS assembler language and might require some minor changes for use
with other assemblers" I would have had no problem and would not have
written ZAS-SLR.DOC.
Two Technical Asides
While I was running the code through the SLR assemblers, I decided to
check my speculation that the libraries contained only 8080-compatible
opcodes (Richard implied repeatedly in the article that the older releases
of the libraries were for 8080 but that the current versions were for Z80
and HD64180 only). SLR's fancier, virtual-memory assembler Z80ASM+ has an
option to flag any non-8080 instructions as errors (its further ability to
flag absolute jumps that could be replaced by relative jumps saved
significant code in VFILER 4.1). I ran it on all modules of all three
libraries. Only four modules contained any Z80-specific opcodes. These
were in the same library-utility modules that have the ZAS-specific ".IN"
pseudo-ops. Since these are the newest routines, they suggest that Richard
is now writing code that will not support 8080- and 8085-based computers.
This in not a criticism; I also think it is time to start taking advantage
of the Z80 opcodes. (The SLR assemblers carry this to a very high degree,
increasing speed and efficiency by making extensive use of alternate and
index registers.) I would leave it to the minority with 8085s to convert
the source and release 8080 versions of the libraries.
While I am on asides, let me note another technical shortcoming of the
coding in the libraries (in ZAS-SLR.DOC I noted a few inconsistencies in
symbol names). None of the routines place their uninitialized data space
into data segments (DSEGs). As a result, COM files made using the libraries
include useless bytes that slow down loading and take up extra disk space.
This has been a problem with all Z programs. It's not an earthshaking
matter, but it is completely unnecessary. VFILER Version 4.1 is the first Z
program (to my knowledge) where the main program has support for DSEGs. I
am now in the process of modifying the libraries code for VFILER. The
linking step is more complicated (except with SLR's virtual-memory linker
SLRNK+, which does it automatically), requiring two linking operations, one
to determine the end of the program space (CSEG) and a second one to perform
the linkage with the data space ORGed immediately after the program space.
With SLRNK this is no big deal, since it will perform two linkages in less
time than other linkers can do one. And during development there is no need
to place the data space precisely at the end of the code space.
The Standard for 8-Bit Assemblers
Now I turn my attention to the part of Richard's ZAS-STD.DOC that I
feel is a real chutzpah. In that file we learn that Richard Conn (King
Richard, perhaps) has decreed a new standard for assembly language that we
all must follow if we want to be a part of the Z programming community. He
argues very eloquently and convincingly for the value of standards, pointing
to the example of Ada, his other major interest.
I, too, place a high value on standards, especially when they are a
help to everyone. But consider how standards are usually arrived at in a
democratic society. Several steps are involved:
1. Notice is given to the community that a standard is being
considered;
2. A committee of experts and interested parties is formed to
oversee the development of the standard;
3. Existing products are examined, studied, and evaluated;
4. Input is solicited from the community at large.
Richard Conn, as far as I can tell, has done none of these things!
First, the announcement of "The ZAS Standard" in ZAS-STD.DOC fell like a
bombshell. I have not seen any notice in Z-News or on Z-Node Central of the
intention to formulate and enforce an assembler standard. There is a BIG
difference between ZAS as the only assembler sold and supported by Echelon
and ZAS as THE STANDARD assembler. Any company may choose to market a
particular product, but only IBM then declares it to be a world standard
(actually, even IBM doesn't do that; their choice just becomes the de facto
standard).
Second, there has been no committee of experts, at least not a public
one. Richard Conn, perhaps with others on the "Echelon Team", seems to have
appointed himself. Does Echelon really think it is the IBM of the 8-bit
world?!
Third, from Richard's own text we see that no examination of existing
products has been conducted. He states plainly in that text that he has
never seen the SLR assemblers. Anyone who keeps himself at all informed
about activities in the computer world other than his own would surely have
noticed the advertisements for the SLR tools. The utterly outrageous (but
true!) claims in those ads should have at least led a standards setter to
investigate.
Fourth, since the community was not even notified of an impending
standard, its input was neither solicited nor given consideration. We have
been presented with a fait accompli. I, and I am sure others, would like to
ask that Richard Conn show some genuine respect for the rest of the 8-bit
programming community by paying a little attention to our ideas in addition
to his own. That means not only listening to us when we talk directly to
him but more importantly looking at the new programs and program
enhancements that we release and at the dialogue on Z-Node Central and the
other Z-Nodes. And it also means actively soliciting input before he makes
decisions that he expects the whole community to follow.
Richard's actions would be open to criticism even if the standard
decreed for us were a good one, but it is not. When Echelon first offered
ZAS, many of us in the community wanted to and tried to support it.
Unfortunately, we found it to be a very mediocre assembler plagued with a
continuing series of problems and deficiencies (I don't want to go into them
here). I have reached the point where I am no longer willing to expend the
effort to check my code for ZAS compatibility; Joe Wright, a member (or
former member?) of the Echelon Team, has even consented to be quoted in the
SLR advertisements (I would, too).
It is with regret that I make the above statements about ZAS. In
general, the Echelon product line is a superb one that lives up to Echelon's
image of excellence. Many of the products (the DSD debugger and REVAS3
disassembler) are, I think, the best available. Joe Wright's auto-install Z
System packages are marvels of ingenuity; ZRDOS offers excellent new
functionality; ZCPR3 itself, the centerpiece, is dazzling. ZAS,
unfortunately, stands out like a sore thumb in this list. When I first used
the SLR assemblers, my immediate reaction was: There are the assemblers that
belong in the Echelon product line!
A Silver Lining
There is a silver lining I am happy to report. Yesterday Richard sent
a message over the INFO-CPM mailing list on the Defense Data Network (ARPA-
NET), where our previous round of files also appeared, noting that he had
just received copies of the SLR assemblers and, though he had not yet used
them, was impressed with them based on the documentation. He also reported
that Echelon and SLR Systems are talking. Perhaps this false start at
standard-setting will be set aright, and we will all benefit.
Final Footnote
Lest anyone take the above remarks excessively seriously, I would like
to temper them by adding that under circumstances such as these I am given
to a degree of overstatement. After all, with nothing but a steady stream
of superb new Z programs week after week, one looks for something to break
the monotony. And why should Frank Gaude' with his Z-News be the only one
to provide some color? Of course, Frank's style and mine are not exactly
the same, and I prefer Cabernet Sauvignon to Zinfandel.