Traits traits

[Brian T. Rice]

nicolas.pelletier3 at free.fr wrote:
> Hello,
> 
> while working on an improved  version of the tutorial, I inspected the
> delegation graph  between Derivable and Cloneable. The  results are in
> the attached  file. There is  a cycle between Derivable  and Cloneable
> which makes Cloneable  a Derivable and Derivable a  Cloneable, so that
> the two concepts collapse.

This is a good analysis. Thanks especially for the diagram to explain 
what the problem is to people.

> The culprit seems to be (as  in the case of True clone) Traits traits,
> which delegates  to Cloneable traits. I  wonder if this  should not be
> considered a  mistake. Maybe Traits  and Traits traits  should instead
> delegate  to a  TraitsDescription object  (imitating what  is  done in
> Smalltalk with Class, Metaclass and ClassDescription).

We might do this, but what we're leaning towards lately is to make 
traits a special slot (like the map is special), so that it is treated 
differently by the lookup procedure (basically only to follow the first 
traits link and none after that). This effectively alters the diagram so 
that traits links are all "directional" (say, vertical where the rest 
are diagonal or horizontal). This would also pretty much make traits 
relationships required, so it would alter the language a bit, but the 
benefit of clarifying inheritance from basic delegation certainly 
outweighs it.

What do you think?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: water.vcf
Type: text/x-vcard
Size: 265 bytes
Desc: not available
Url : /archives/slate/attachments/20040923/764e6b41/water.vcf