B 5000, MCP, etc - was Re: And speaking of ALGOL

Nigel Williams nw at retrocomputingtasmania.com
Thu Aug 13 03:25:17 CDT 2015


> On 13 Aug 2015, at 1:04 pm, Toby Thain <toby at telegraphics.com.au> wrote:
> For more on this, see "Classic Operating Systems," Per Brinch Hansen, which reprints the paper "Operating System for the B 5000", Clark Oliphint (1964).
> 
>  "Two of the major B 5000 design objectives were (1) that all programming was to be done in ALGOL and COBOL, and (2) that the operation of the system was to be directed by a Master Control Program (MCP) which would relieve the operator and---especially---the programmer of virtually all the inefficient and error-causing details of peripheral unit designation, memory area assignment, and so on. The simultaneous and coordinated design of the computer and the programming system has produced a hardware-software system so well integrated that all B 5000 users employ the standard programming system (with minor modifications for special applications in a few cases). It has been the experience of B 5000 users that the exclusive use of compiler languages in programming gives advantages in documentation, program preparation, and debugging which cannot be over emphasized. ..."

Oliphint provides an interesting interpretation of the “design objectives” but as we see in the oral history (at UMN.edu) some of the decisions were arguably made (or at least supported) for different reasons, even within the groups who made them. For example, some people within the B5000 design team were certain that ALGOL was the better language and would dominate scientific computing (it wasn’t clear to them that FORTRAN would last the distance). They supported this view by referencing CACM’s decision to publish algorithms solely in ALGOL. Others within the group saw risk in an ALGOL-only machine and fearing a too restricted customer base urged support for COBOL to be added (hence the haphazard attempt at a set of string operators that was later completely replaced with a new design for the B6700 - many ideas from the B5500 were evolved for the B6700 but the string operators were not). The feeling was the B5000 never really delivered on the promises, and it was only when the B5500 was released, and after several iterations of the MCP (the operating system) did it really deliver the benefits cited for many of the original design decisions.




More information about the cctalk mailing list