On 22 February 2014 18:35, John Ehrman <[email protected]> wrote: > As I remember, the oldest ancestor of PL/X was BSL, "Basic System(s?) > Language". A talk on it was given at SHARE by Bob Brittenham, who led the > team that developed it. Back in those days IBM shipped many modules with > source code, and in several cases the "shipper" forgot to include only the > "clean" generated assembler source and included a version with the embedded > BSL statements.
I think this was the norm in most cases in the early post-BSL days - certainly for mainstream MVS code listings on microfiche, and in the optional source that was generally orderable. I can remember complaining to IBM when they shipped PL/S-generated assembler without the PL/S statements included, and on at least one occasion they corrected it. IBM even produced a manual "Guide to PL/S-Generated Listings", GC28-6786, and a follow on "PL/S II Guide", GC28-6794 that explained some aspects of the way the compilers generate code, assign compiler-generated identifiers, and so on. > Gio Widerhold (of Stanford Med Center's ACME time-sharing > system -- written in PL/I, as I remember) and I (at the Stanford Linear > Accelerator Center at the time) published a paper in SIGPLAN Notices called > "Inferred Syntax and Semantics of BSL". Things tightened up quickly after > that, and much less was said externally about BSL and its successors, the > first of which I believe was PL/S. There were a number of efforts to produce compatible PL/S compilers even in the absense of formal language specs, the most notorious of which was the Rand Corporation's RL/S. http://www.mxg.com/thebuttonman/html/button196.htm http://www.mxg.com/thebuttonman/html/button213.htm I am aware of at least two unpublished research efforts toward this end that either didn't get very far, or were discouraged by IBM's response to Rand. Plenty has been written on this matter. And IBM uses internally a PL/S-like language PL.8, which arguably (and again in the absense of full documention) is a much better language than C for low level coding for things like millicode and device drivers. Although PL.8 is a modification to the GPL-licensed GCC, sadly IBM has chosen not to make it available, though they have published much of the compiler middle and back end updates for s390x which also enhance C and any other GCC front end performance and capabilities. Tony H.
