Re: How to get Ada to “cross the chasm”?

["Dan'l Miller" <optikos@verizon.net>]

On Wednesday, April 18, 2018 at 12:02:18 PM UTC-5, Lucretia wrote:
> ...
> There are a lot of people who would love to see a new Ada version break away from the previous
> versions, possibly even a rewrite removing all the crap that's not really required and basically, an Ada for
> the 21st century.

This would be extraordinarily difficult to do •within• the Ada community/standardization, because doing so would of course be a critique of Ichbiah's decisions and thus a disloyalty to Ichbiah himself.  This would be far easier to do in a new (reference-implementation) compiler that, while optionally providing an Ada backwards-compatibility mode, feels free to go explore a new era, unfettered by, say, Ada83.  In particular, the compiler could be modular in the sense of unplugging the Ichbiah semantics of compiler-enforced rule-set and plugging in, say, Rust's borrow-checker/et.al. on a per-project basis.

There is Ada-the-syntax that (with the exception of discriminants) is awesome.  Then there is Ada-the-Ichbiah-compiler-enforced-ruleset-from-Ada83 and then there is Ada-the-1995-compiler-enforced-ruleset and then there is the Ada-the-2005/2012-compiler-enforced-ruleset.  We have 3 or so eras of definition of safety.  Rust shows a 4th.  Randy is proposing some sort of 5th definition of safety for provably-correct tasking in Ada2020.  With so many schools of thought vying for the safety crown, this area is ripe for pluggable modularity for different schools of thought.  Even better if the axiom systems were made not-mutually-exclusive:  I'll take no-Ichbiah, some of Tucker's, all of Randy's, and the borrow checker from Rust—or some other combination for a different project.  Hence, why the compiler-enforcement of rule-set (and the ARM that defines it) should be some form of source code that declares an axiomatic system of logic, vaguely similar to Zed Notation.  Now •that• would be an Ada for the 21st century.

> ...
> Yet people still bitch about how "hard" it is to get an Ada program to compile.

Make it harder.  Let the programmers-love-to-typewrite-me languages continue conquering the ease-of-compiling marketspace.  Swift, C#, and Java value this over almost all other goals, especially Swift's seemingly never ending renaming to even more-beautiful identifiers and typing-convenience features.  Ada needs to fully conquer the source-code-is-a-rigorous-mathematical-proof marketspace; Rust and perhaps Ocaml+Coq are gaining on Ada in that race.

> ...
> This is where Ada really shines, but it could go further with endian describing. There is no other
> language which does this, at all

Well, there was PL/I from which Ada got the idea of meticulous control of the layout of records.  Ichbiah's Green was competing in a 1970s ecosystem that had a background commentary of “why not just pick PL/I”.  Multics had picked PL/I.  Prime Computer had picked PL/I, refashioned as PL/P.  Intel and Digital Research had picked PL/I, refashioned as PL/M.

https://www.ibm.com/support/knowledgecenter/en/SSY2V3_4.2.0/com.ibm.entpli.doc_4.2/lr/lsh-bigendian.htm