This is not my idea but the ones that nathanael shares, builds and 
throws at/with/on us :).
If you could see our black boards they are full of jumping ideas.

Basically we can associate an interface (or multiple ones) that specify 
how a trait method
can be accessed. By default the interface would be the complete trait 
methods. This would
support private methods in traits but not by introducing another gory 
construct. The clients
could also see different interfaces. This mechanism would support the 
expression of privacy in the freedom way of smalltalk. In fact what 
andreas is looking for Morph.


On Monday, February 10, 2003, at 10:45 PM, Withers, Robert wrote:

> Ok, Stephane, please all'y'all :) join the squeake mailing list and 
> share
> with us your ideas.  We just went through a discussion about how to 
> mix the
> openness and malleability of Smalltalk with the security of E.

For that I would really like to find the time to play with mirrors. But 
I don't :(((



> Result:
> Squeak will host the SqueakE language and use its tools to reflect into
> SqueakE  (at least until we can build a refractor system - courtesy of 
> Mark
> Miller).
>
>
>> -----Original Message-----
>> From: Stephane Ducasse [mailto:ducasse at iam.unibe.ch]
>
>> hehe :) the power of pointers pointing the the same things and traits
>> to not have state so
>> no offset mess.
>
> Good thing there is no state in shared traits, I'm guessing.  :)
>
>> yes the only thing is that when sharing compiled method is
>> that we have
>> to copy the ones
>> that contains a super reference. But this a smart trick from
>> nathanael.
>
> ahh, that makes sense.  So you compile in context...    But, does this 
> mean
> that methods from a Trait cannot access instance variables directly?
> (semi-rhetorical)  I must look at the context you compile in.  I guess 
> it
> would be the defining Trait.
>
>
>>> I can't wait to hear about the other goodies coming from
>> Bern...  ;-)
>>
>> ClassBoxes :) (Modules with local rebinding) soon we will
>> have a public
>> paper and alex is reimplementing it influenced by the
>> implementation of
>> Steven Pair for multiple namespaces.
>
> awesome!  (I think - it sounds cool!)   Does this get into the Oasis 
> realm
> of plugging a module into a backplane to resolve references?  What has
> happened to Les, btw?  The oasis site is down.
>
>
>> A smalltalk without direct access to instance variables.
>> That's an non
>> conceptual one :)
>
> oh, yes!  But what about assignment?  We could definitely use this in
> squeake.
>
>
>> Open privacy for traits. Quite cool but not yet done....
>
> This also sounds similar in goals to squeake.
>
>
>> at the end we hope to have our own language :) possibly with
>> scripting
>> facilities, reflection, mirror and modules. We will get there I hope
>> this is too exciting.
>
> Well this is definitely the plan for squeake.  It must be a different
> language, although we are seeing if we could in fact support the ANSI 
> spec.
> The fundamental agreement was that it must have Squeak's openness and
> malleability - that is the real power of Smalltalk.
>
> cheers!
> rob
>
>
Prof. Dr. Stéphane DUCASSE (ducasse at iam.unibe.ch) 
http://www.iam.unibe.ch/~ducasse/
  "if you knew today was your last day on earth, what would you do
  different? ... especially if, by doing something different, today
  might not be your last day on earth" Calvin&Hobbes