Where should the type information be 166

Stephen Fuld

PL-1 on VAX-VMS was a very fine compiler; I was surprised at how well it worked for systems programming. I was involved with a project to support a Xerox 9700 laserprinter on VMS. The project replaced sixty-some regular line printers in DEC's Marlboro engineering facility with two large laser printers.

The Xerox printers had an embedded PDP-11 (with RT-11 as the MS-DOS equivalent: Really just used as the second-stage boot loader). There were two ways to speak to them: Xerox Star office clusters with send a proprietary page description language over ethernet (using XNS protocol stack). This format was very poorly described. IBM mainframes would connect to an IBM channel controller emulating a 1402N1 printer, but with escapes to a command set that allowed some graphics, programmable fonts, etc. This interface was well documented.

I was working for ACC (buttociated Computer Consultants at the time), when DEC approached us to make them an IBM channel emulator to plug into the Unibus. I wrote the firmware for the Z80 that was embedded in the controller, as well as the VMS device driver. The result was a reasonable selector channel emulation that put very little load on the VAX. Paul Kyzivat from DEC wrote the print spooler in PL-1. Later DEC let us have the spooler and give it to other VMS installations that wanted the same thing, and I got a chance to look at the code when we were contemplating adding a few more features.

PL-1's ON condition handlers plugged directly into the VMS exception handling mechanisms, and the standard calling sequence allowed mixing of any languages for routines to do special things if you liked. The standard calling sequnece represented all arguments by descriptors that included a type, a length and an address. C was the ONLY language unable to make use of the standard calling sequence.

If I remember correctly, this was around 1984.

--