🚀 go-pugleaf

Andy Leighton <an...@azaal.plus.com> wrote:
> On Sat, 23 Jun 2018 11:47:34 -0400, Peter Flass <peter...@yahoo.com> wrote:
>> Andy Leighton <an...@azaal.plus.com> wrote:
>>> On Sat, 23 Jun 2018 07:04:25 -0400, Peter Flass <peter...@yahoo.com> wrote:
>>>> J. Clarke <jclarke...@gmail.com> wrote:
>>>>> On Fri, 22 Jun 2018 13:56:09 -0400, Peter Flass
>>>>> <peter...@yahoo.com> wrote:
>>>>> 
>>>>>> Gareth's Downstairs Computer
>>>>>> <headstone255.but.n...@yahoo.com> wrote:
>>>>>>> On 22/06/2018 11:37, Peter Flass wrote:
>>>>>>>> 
>>>>>>>> Isn't Mono a clone of .net? What does the common runtime do especially that
>>>>>>>> other libraries don't? (actual question, not rhetorical)
>>>>>>> All the source languages link together.
>>>>>>> 
>>>>>> 
>>>>>> Sort of like all the gnu languages, the IBM mainframe languages with LE, or
>>>>>> all the DEC languages with the common runtime? Leave it to microsoft to
>>>>>> claim they invented sunshine.
>>>>> 
>>>>> Except that the gnu languages and LE don't produce binaries that will
>>>>> run without further adjustment on an Alpha, an Itanium, a Sparc, and a
>>>>> PC, which, in principle, is true of CLR (I say "in principle" because
>>>>> I don't know if the versions of Windows that made it to the Alpha and
>>>>> Sparc are recent enough to support CLR).
>>>> 
>>>> In theory they can, although I don't know if any do in practice. ELF
>>>> binaries can contain objects for multiple architectures. Or are you talking
>>>> about something that's interpreted? That sounds like a spectacular waste if
>>>> resources.
>>> 
>>> Not interpreted but just in time compilation.
>>> 
>>> Normally with .net the compilation stage runs ngen and compiles the CIL
>>> (common intermediate language) down to native ahead of time and people
>>> deploy that. So it isn't usually the case that you can take the x86
>>> version of a .NET application and run it on another architecture. You can
>>> however choose to take the CIL assembly and run that anywhere there is a CLR 
>>> (and appropriate support libraries which is where Mono is sometimes lacking).
>>> I would also note that there are some cases where the JIT version is faster
>>> than the ahead of time compiled ngen output.
>> 
>> As someone said, it sounds like Java, where the bytecode can be generated
>> from any language if the compiler is set up to handle it.
> 
> Yep pretty much. Both the CLR and JVM are stack-based virtual machines
> with a pretty decent JIT (although I think the Java JIT is a bit better).
> Java doesn't have AOT (ahead of time) compilation, .NET can have. The
> JVM is more of a classic VM in approach. The CLR has support for lots
> of stuff - like tail recursion, pointers, new value types and can deal
> with generics better.  Some of that is due to CLR being implemented 
> later, some more to do with philosophy. CLR doesn't count as something
> innovative in my eye.
> 
> 

But, as usual, M$ did something different just to be different (and
incompatible). Why follow an existing standard (real or de-facto) when you
can define your own "standard"?

-- 
Pete