Re: Could you write a BSD like os in ADA?

[rieachus@comcast.net]

On Monday, August 22, 2016 at 8:11:36 AM UTC-4, Dmitry A. Kazakov wrote:
> On 22/08/2016 13:38, J-P. Rosen wrote:
> > Le 22/08/2016 à 10:14, Dmitry A. Kazakov a écrit :
> >> But designing a new OS I would consider doing it really new. No I/O, no
> >> files, only memory mapped persistent properly typed objects.
> > Multics?
> 
> I don't think Multics had typed persistent objects, especially 
> user-defined ones. You cannot stull all possible types into the OS. It 
> must be open-end.

Um, technically, Multics had three types of files.  Boot tapes (yes the 1/2" wide tape on 12" or smaller spools), backup tapes (managed by Multics, if you deleted something you didn't mean to, go into the backup system and it should be there) and finally multi-segment files which were a way to combine segments when you needed more than a megabyte.  However, to most users all of this was irrelevant.

What is really relevant to this group is that Ada would not have existed without Multics (and e-mail).  The Ada language was developed on several Multics machines, and as a result all Ada developers knew the Multics way of thinking.  The most important thing (IMHO) that Ada got from Multics was the idea that the OS or compiler should do all the work of maintaining a consistent source tree.  I never got used to linkers, writing make files, etc.  (Multics did not require linking.  If you ran an executable, and it called another unit it would dynamically link it in. If it didn't exist?  Multics would tell you and allow you to write it, compile it, resume your program and it would use the unit you just created.)

Why wasn't there an Ada compiler for Multics?  At Honeywell Small Systems (and other names) where I worked we were tracking the development of Ada with a compiler that ran on Multics and generated code for DPS6 and other small systems.  The problem was that our compiler was intended as a systems development tool.  As a result it could ride roughshod over the OS.  Not only did it "know" how to access the OS internals, it was used to develop things like the e-mail system.  All OS internal calls could be made from within Ada/SIL (for systems implementation language).

We arranged for Dansk Datamatik to port their compiler to our system, and we validated it.  Large systems decided to also port the DDC compiler, but to GCOS-8, not Multics.  I never found out if they finished, but it was years late...