🚀 go-pugleaf

chl@cs.man.ac.uk (Charles Lindsey) writes:

>In fact, I understand that the ALGOL 68RS system does in fact implement the
>Bekic proposal, or something close to it.

Only one of the ALGOL 68RS implementations, on the RSRE Flex
architecture, permits procedures without the normal scope restrictions.
I don't know of the Bekic proposal; the implementation on Flex
uses normal ALGOL 68 syntax, only the scope rules being relaxed.

The Flex architecture, designed around 1978, is geared towards
efficient implementation by micro-code, as on a Perq.  Heap activation
records are always used, and retained by 'shaky pointers' on exit to be
reused at the next call of that procedure.  The (micro-coded) garbage
collector recovers these retained activation records.

Procedure values are used as capabilities on Flex, the microcode
ensures procedures (and pointers) cannot be fabricated.  This provides
a secure multi-process environment in a single address space, with an
extensible operating system.

Flex also has a persistent store, so these first class procedures can
even live on disc when you turn the power off!  In fact the only real
difference between a procedure you can produce and an operating system
procedure is that the operating system procedure originated from a boot
floppy.

It's a shame Flex did not receive more publicity, as it implemented
features much discussed later.  I talk about it in the present, as I
have a Perq running Flex next to me, but it doesn't get switched on too
often these days.
--
Richard Wendland			richard@praxis.co.uk