NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/unix/aix/8193

Internet Message Format | 1992-07-26 | 6.7 KB

Path: sparky!uunet!mcsun!Germany.EU.net!gmd.de!jvnc.net!yale.edu!spool.mu.edu!umn.edu!csus.edu!netcomsv!mbeckman!mbeckman
From: [e-mail] (Mel Beckman)
Newsgroups: comp.unix.aix
Subject: Re: System36 Emulator
Message-ID: <[e-mail]>
Date: 26 Jul 92 19:37:57 GMT
Article-I.D.: mbeckman.01050810.9gfok1
Reply-To: [e-mail]
Organization: Beckman Software Engineering
Lines: 119
X-Mailer: uAccess LITE - Macintosh Release: 1.5v5


Ian Graham said:
> Mr Beckman is quite right in describing RS/36 as a complete runtime environment
> emulating the System/36 and executing original System/36 binaries.  (You may
> however wish to think carefully about license infringement when executing
> binaries for which you you do not own the source rights.)
>

You don't need source rights to execute binaries for
packages to which you currently have object rights. If you have
the "object" (binary) rights to the S/36 software you run, you can execute
that binary on any equipment you wish. I'm not taking a position one
way or the other on the overall desirability of either RS/36 or UNIBOL,
but it seems to me that it is UNIBOL, which requires recompiling source
programs, where users must be concerned about owning source rights (and
having the source code).

  Ian said>>The performance of the machine
allows RS/36 to be faster than a System/36 but because it is a ** native **
implementation, UNIBOL provides considerably faster performance still !!<<

 Tests conducted by NEWS 3X/400 magazine showed UNIBOL only slightly
faster than RS/36 in some situations. Still, I wouldn't say one way or
another one product is faster, because our tests were simply
functional checks, not well-designed benchmarks (we point this out in
the reviews). I don't know if Ian has been able to run such benchmarks
on RS/36, but lacking such a formal test, nobody can say which
is faster, if in fact there is a significant speed difference at all.

As to the "native" claim, our understanding from extensive
discussions with ICS engineers is that UNIBOL compiles RPG programs into an
intermediate language which is then interpreted by a runtime system. Our
definition of native is compilation to machine code that is then directly
executed. I see no inherent qualitative advantage in UNIBOL's approach
over RS/36's interpretation of S/36 machine instructions (or vice versa).
Further, most RPG programs are I/O bound, not compute bound, so the speed
of the I/O systems is what controls overall throughput, not instruction
execution speed. Our tests also bore this out. I do believe that S/36
machine code emulation simplifies RS/36's job of replicating the many
quirks of S/36 RPG behavior. ICS had to identify and replicate these quirks
by hand; RS/36 gets such quirks as a free biproduct of low-level emulation.

   ICS tells us that they consider interpretive execution to be "native",
but I've not yet heard a convincing argument from them to make this case
(but I'm still open to convincing).

 Ian said>>UNIBOL provides complete support for OCL, screens (SDA), help, messages,
autoreport, editors, RPGII, COBOL, RPG400 (including ASNA and BPS), POP, any
ascii/ansi terminal (up to 9 sessions per terminal), 5250 devices, report
generator and full capability at all levels to interface to other native AIX
applications and utilities.  <<

  While both products make claims to provide these features, their
success varies, with no clear winner. For example, most S/36 users likely will
be more comfortable with the RS/36 POP implementation, a virtual clone of that
IBM tool. RS/36's POP also includes a clone of the FSEDIT full-screen editor;
UNIBOL doesn't.

On the other hand, UNIBOL's POP replacement, which more closely resembles
AS/400 PDM, probably is more functional for users with AS/400 background.
However, the lack of a midrange-style fullscreen editor is a significant
problem. UNIBOL supplies a clone of IBM's circa-1983 SEU line editor,
or you can use the Unix vi editor. While vi is full-screen, it's not very
good for editing RPG programs, which are column oriented.
S/36 users definitely have a longer learning curve
with UNIBOL's tools than RS/36's.

Similar tradeoffs apply to all the other support features (terminal emulation,
SDA, report generators, AIX access). I see no clear overall winner in any
of these categories; users will have to evaluate these areas to see which
match their own preferences.

Ian said>>Compilation speeds of 130 times faster than a 5360
make it a dream of a development environment.<<

One reason for such fast compilation speeds is because UNIBOL's RPG
compiler produces an intermediate language, not optimized RS/6000
machine code. UNIBOL would compile much more slowly (and require much more
memory for compilation) were it to go directly to RS/6000 machine code,
or even if it produced C code for subsequent compilation by IBM's optimizing
C compiler.  Neither RS/36 nor UNIBOL provide such options (although both
vendors may be working on them).


 Ian says >>Not only does it provide
a replacement System/36 with vastly improved performance, it allows users
(currently 2000+) to retain and build on existing experience while at the same
time take advantage of a wealth of new software languages and applications.<<

   In our discussions with ICS computing, we were told that the "2000+"
user count applies to the entire installed user base of UNIBOL systems, *not*
to RS/6000 UNIBOL sites. ICS has never has given us a firm figure on RS/6000
installs, but last fall said "less than 200" when we asked for a ballpark
figure.

   Clearly the broad range of systems which run UNIBOL speaks well of
its track record in Unix shops in general, but the products are not
especially compatible with each other, and are maintained by separate
technical groups. UNIBOL on the RS/6000 has its own high and low
points independant of other UNIBOL implementations; potential users
should evaluate these points apart from these other implementations.

   If anyone would like copies of both the RS/36 and UNIBOL reviews, drop
me your postal address via Email.  The August 1992 issue of NEWS 3X/400 magazine
revisits this issue with updated product information.

  Mel Beckman
  Senior Technical Editor
  NEWS 3X/400 Magazine

_____________________________________________________________________
| Mel beckman                  |    Internet: [e-mail] |
| Beckman Software Engineering |  Compuserve: 75226,2257            |
| 1201 Nilgai Place            |       Voice: 805/647-1641          |
| Ventura, CA 93003            |         Fax: [phone]          |
|______________________________|____________________________________|
 "Internet is big.  Really Big.  It gives the idea of
  infinity much better than infinity itself."
   (with apologies to Douglas Adams)