| Chuck Stevens 2004-08-30, 3:55 pm |
| "Clark F. Morris, Jr." <cfmtech@istar.ca> wrote in message
news:2peee9Fiva50U2@uni-berlin.de...
> Chuck Stevens wrote:
> There were two commercial systems, the 705 - a 6 bit character machine
> and the 707 - a 10 decimal digit word machine. Apparently IBM thought
> that they could move the 705 people to the 707 follow-on 707x series
> (transistorized) because apparently more 707's were sold than 705's. I
> am still trying to figure out why the follow-on was a 7080 and not a 7050.
Thanks for the clarification; I was unaware of the IBM 707. But I think the
705/7080 was also a fixed-word machine -- I don't remember the word size,
but 5 characters (30 bits) seems to be drifting around my memory. I also
remember a large rotary accumulator (360 digits?) as part of the design. I
taught classes in 705/I/II/III architecture, but that was, after all, 35
years ago ...
> My impression in the 360 - z series area is that at best the platform is
> not suffering a net loss of customers. My impression from reading
> Unisys ads on Clear Path is that either or both major Unisys
> architectures are being emulated on Intel.
As far as I know the 2200 architecture is directly supported by custom
hardware throughout the line.
So also is the upper end of the MCP architecture (the ClearPathPlus Libra
590, 580 and 185) whereas the midrange and smaller members of the MCP line
are based on the MCPvm environment (the ClearPathPlus Libra 520, CS7201,
LX7100 series, LX6000 series, and the LX140 laptop).
> My general impression is
> that the overwhelming customer base is the existing one which is
> shrinking.
I don't have access to the sales figures, but I do recall the comment from
our management a year or to ago that more MCP-based systems were sold in the
previous fiscal year *than in any year in the history of the Burroughs Large
System architecture*. I will grant that there seems to be some "bad press"
about non-commodity-based enterprise servers, but there are a fair number of
companies out there who feel the features offered by such environments --
security, reliability, scalability, recoverability -- are superior.
> I am assuming
> that the major Unisys architectures are built on the 48 bit word and the
> 36 bit word.
Certainly correct for the MCP systems, and I believe correct for the OS2200.
Both have extensive support within the architecture for eight-bit bytes.
> The MCP side also had 6 bit character oriented
> instructions back in the 1960's when I last looked at it and don't know
> if hardware support was added for the 8 bit bytes.
Can't speak to the 2200 (and your comment wasn't about that system anyway),
but 8-bit support has been around as an option at least for a *very* long
time (a B6500 COBOL-60 reference manual describes DISPLAY-2 as 8-bit
EBCDIC).
In COBOL(68) on the B6700 and A Series, USAGE DISPLAY was 8-bit EBCDIC;
USAGE DISPLAY-1 was 6-bit BCL data. Manufacture of machines capable of
manipulating 6-bit data ceased about twenty years ago, and the DISPLAY-1
USAGE and related constructs were deleted from the COBOL compiler and
related documentation in 1987.
A quick glance through the patch history seems to confirm my suspicion that
support for 6-bit BCL data was never provided by COBOL74 (to say nothing of
COBOL85) at all.
-Chuck Stevens
|