On Sunday 06 October 2002 04:20 am, Stephane Ducasse wrote:
> Hi Brian
>
> If you want to see what we are doing with Squeak and Traits look at
> the first (not that good paper)
> we are writing a good one.
> http://www.iam.unibe.ch/~schaerli/research/index.htm
>
> Thanks for the advice
> Stef

Cool, I like this ALOT. Is there a chance we will see this in Squeak?

It jives with my observation that Squeak is a protocol-typed languaged 
rather than a typed-language. There is probably a OOP term for what I 
am saying, but I don't know it... 

As long as an object understands a method, it works ( well, as long as 
the message does what you want ). 

I should qualify this more and say, that what matters more often in ST 
is NOT that a object understands a method, but that it understands a 
PROTOCOL, a group of messages that get things done. So thinking it 
understands the complete protocol can bite one on the rear. The 
Collection classes are a perfect example of this. The all support 
different junks of what is currently thought of as one protocol.

With traits, we could divide these out these out into seperate traits, 
and retain the collection heirarchy to preserve reuse of state and glue 
code. Then plug in the traits. Once done, a browser could show that 
"This class supports the sorted trait, this one doesn't, etc etc".

Something like traits makes this very possible. A trait browser would 
let us see what objects easily understand a given set of protocols.

Also, it's a WAAAY better idea than JAVA's empty interface classes. 
These guys can actually have some meat on them.

Is there a chance we'll see Traits in Squeak?  <:)

This is a really really really cool idea...

-Daniel