Re: Adacore joins with Ferrous Systems to support Rust

["Randy Brukardt" <randy@rrsoftware.com>]

"Luke A. Guest" <laguest@archeia.com> wrote in message 
news:stgj3j$12g8$1@gioia.aioe.org...
> On 03/02/2022 11:30, Simon Wright wrote:
>> "Luke A. Guest" <laguest@archeia.com> writes:
>>
>>> At uni, we didn't even touch on tagged types, but used controlled
>>> types
>>
>> Of course controlled types are implemented as tagged types. ISTR Dmitry
>
> We had a book with a reading guide. We weren't shown the specs showing the 
> type of controlled/limited_controlled. So, when you don't have that 
> information and only know that you can derive from it and then implement 2 
> or 3 subprograms, you don't need to know about OOP in Ada.
>
> Just like we didn't touch generic packages until much later in the year, 
> generic subprograms, yeah, because u_d was required early on.
>
>> objecting to this, I think he was right (philosophically; I've no idea
>> about the practicality of alternate approaches).
>
> How else would you do controlled types?

Ada 9x originally had a bunch of magic attributes (similar to streaming). It 
was very complex and got dumped in the dustbin during "scope reduction". 
Later on, some of us were bemoaning that a critical feature (finalization) 
had gotten lost in Ada 9x, and Tucker came up with the idea to build it on 
top of tagged types as a massive simplification (at the loss of a bit of 
power). People often complain that Ada finalization is complex, and it is, 
except all of the alternatives are a lot more complex. :-)

                   Randy.