On Thu, Dec 17, 2020 at 11:35 AM Bakul Shah <bakul@iitbombay.org> wrote:
Funny how we seem to rehash the same things over the years!

Ars longa, vita brevis.
In a 1988 comp.lang.misc thread when I expressed hope that "a major
subset of Algol 68 with a new and concise syntax (sort of like C's)
can make a very elegant, type safe and well rounded language.",

Thanks. The URL is <https://groups.google.com/u/2/g/comp.lang.misc/c/qkmB_3zuC7Y/m/-Bk9z-oZaqYJ>.
Piet
van Oostrum[1] commented the combination of dynamic arrays *and*
unions forced the use of GC in Algol68. Either feature by themselves
wouldn't have required GC!

I can't find this anywhere in the thread or elsewhere in comp.lang.misc. Do you have a reference?

Here's the link: https://groups.google.com/g/comp.lang.misc/c/qkmB_3zuC7Y/m/erN_TfDF38IJ

In any case, I don't understand how a safe language with pointers can avoid the need for *some* kind of GC.

Note that I was talking about a "type safe" subset as opposed to as stricter "safe" subset.

One can argue that Pascal was type safe even with new() / dispose() for pointers.

[Referencing a dead pointer is a runtime error, just like divide by zero]

I think his point was more that each of dynamic arrays by themselves wouldn't require

GC and the same for unions. Remember I was talking about a vague typesafe subset.

So the question is which features to remove? Rmoving pointers would certainly be under

consideration.

[My exposure to Algol68 was when I had stumbled upon Brailsford and
Walker's wonderful "Introductory Algol 68 programming"

Alas, I can only find this at one shady site, or in hardback at Amazon for USD 20 which is a lot for a pig in a poke (no preview).