Well according to the author of RTalk , he never intended for this to be a full implementation of smalltalk , but later in his video presentation mentions that RTalk ended being a full implementation of DigiTalk which is the smalltalk his company was using for its legacy code. He actually does special mentions how he implemented non local returns. He even claims that not only he was able to replicate all digitalk bytecode but even instruct JVM to construct bytecode unrelated to Java.

But me being a total beginner with smalltalk means I could as well misunderstood his presentations. Of course without having an opportunity to take a look at the actual code this is just a moot point.

----- Original Message -----
From: Camillo Bruni <camillobruni@gmail.com>
To: Pharo-project@lists.gforge.inria.fr
Cc:
Sent: Tuesday, 30 October 2012, 23:18
Subject: Re: [Pharo-project] [OT] RTalk

On 2012-10-30, at 18:39, Jan Vrany <jan.vrany@fit.cvut.cz> wrote:

> On 30/10/12 15:39, dimitris chloupis wrote:
>> Hey Torsten and thanks for the reply.
>>
>> I am not here to recommend pharo to move to JVM. Rtalk author also
>> mention this, moving a language to another platform like JVM is quite
>> easy ,
>
> I've spent quite some time playing with Java and JVM. I don't think in case of Smalltalk language it would be exactly 'quite easy' to move
> to JVM.
> There are certain constructs, prevalent in Smalltalk codebase, that would be tricky to implement *efficiently* on JVM. Non-local returns,
> proceedable exceptions, class extensions, snapshotting to name some.

not to mention instance migration! (which might be circumvented in non-development mode)