[Info-vax] open source OpenVMS (Re: Oracle-RDB seminar notes)

[Richard B. Gilbert]

On 3/31/2013 12:22 PM, Johnny Billquist wrote:
> On 2013-03-31 17:08, Bill Gunshannon wrote:
>> In article <VOudndrZt_ZD4MrMnZ2dnUVZ_jqdnZ2d at giganews.com>,
>>     "Richard B. Gilbert" <rgilbert88 at comcast.net> writes:
>>> An error in the Payroll program is just a huge nuisance!  If an
>>> application error can kill someone, it's more than a mere nuisance!
>>>
>>
>> Oh yes, Ada is so much safer.  Let's see....
>>
>> Task:  Arm Bomb, open bomb bay doors, drop bomb.
>>
>> Develoment systems does these three in parallel simultaneously.
>> Final system does them asynchronously:
>>                                Arm Bomb, drop bomb, open bomb bay
>> doors.   :-)
>>
>> Anybody who designes a language that has any ambiguity built into it
>> has never heard of Murphy.
>
> Speaking as someone who have actually done military (and other) stuff in
> Ada, I think this is a little bit of crock.
>
> Yes, *any* language have ambiguities. It's amazing what weird case pop
> up, which exposes undefined things in languages.
>
> But the kind of scenario you talk about above, Bill, is something that
> you should always explicitly serialize in a program. If you ever write a
> program where such action is serialized implicitly by the programming
> language is just plain asking for things to break. In Ada, don't get
> things parallelized by random. It happens if you have several tasks (and
> yes, a task is a type in Ada). And if you have independent tasks, of
> course things can happen in parallel. It's more of the odd situation
> with some run time systems where it might appear to happen in serial.
>
> Ada definitely have both strengths and weaknesses, just like any
> language. The fascistic type system, and the design in two parts, where
> you declare your external interfaces separate from the code in a way
> that really encourages clean separation between modules is something
> that really helps when you have clowns that write code. There is very
> little (close to none) of the type of errors that manage to get by the
> compiler, which are types of errors that just overflow C.
>
> However, the same fascist type system can really be a headache when you
> know what you want to do, and you know all the issues and you know that
> they don't really matter, but you still have to convince or work around
> the compiler.
>
> Would I write an operating system i Ada? Probably not. But for the
> reasons that have been mentioned here so far, which seems to be more of
> the uninformed and preconceived form. But then again, I also happen to
> like writing all my code in Macro-11...
>
> Ada works just fine, and some things are very easy to do in Ada, which
> is a lot more headache in some other languages.
>
> Gnat is not (I think) a good measure stick for the language. I know that
> it's the only implementation most people have been exposed to, and some
> people might even know the innards of. However, last I looked, Gnat was
> not even a properly certified Ada compiler, and no matter what some
> people say, I don't think the gcc backend is the best tool for doing a
> compiler. And finally, with Ada, the runtime system is important but
> complex. It sounds like that part in Gnat might be problematic.
>
> Ada the language prevents a lot of crap written by some people to even
> get by the compiler. But in the end, nothing will help you if you are
> not competent to write a program in the first place. No compiler will
> catch logic errors. (Which is, by the way, also one of my main points
> against strongly typed languages. It gives you a false sense of
> security. Crap programmers will still produce crap, and good programmers
> don't have problems with types.)
>
>      Johnny
>

I'd say it's easy to write code.  If you insist that the code must solve 
someone's problem(S), it gets much harder.  If the solution(s) must be 
correct, it gets still harder!

:-)