Joseph Kiniry:
> What I never completely understood is why ANY is not just the empty
> dual of NONE; i.e., the top of the behavioral (conforming) inheritance
> hierarchy.
>
> If this were the case, we could then refactor the existing features of
> ANY into appropriate mixin classes, and make all classes that
> behaviorally inherit from ANY today behaviorally inherit from the
> appropriate classes in this (small) set of new classes. New classes
> can choose to use behavioral or non-conforming inheritance as they
> wish, but all classes conform to ANY.
Seems fine, but perhaps too disruptive to existing code. As you
correctly diagnosed though, it is a strawman when it comes to
the SE2 repeated inheritance rule, and moving stuff out of ANY
will just move the problem to the new classes where you put
the things which are now in ANY.
> Has anyone actually written up the type rules for all of these new
> proposals flying about?
You are probably the only known Eiffelist who is fluent in
type theory, so if you haven't written them up, it's unlikely
someone else has unfortunately.
The SE 2 rule on repeated inheritance should be straightforwardly
typesafe, in that transformation of a type theory that describes ETL3
Eiffel into one that describes SE2 Eiffel -- with regard to this
difference only -- should only involve removing some rules, and
not adding or changing any, so the change should conserve any
correctness properties of ETL3 Eiffel.
------------------------ Yahoo! Groups Sponsor --------------------~-->
Make a clean sweep of pop-up ads. Yahoo! Companion Toolbar.
Now with Pop-Up Blocker. Get it for free!
http://us.click.yahoo.com/L5YrjA/eSIIAA/yQLSAA/saFolB/TM
--------------------------------------------------------------------~->
---------------------------
http://www.eiffel-nice.org/
--------------------------
|
Try Searching:
servers, voip, java, networking, microsoft ...
|
|
|
| 
