[erlang-questions] Theoretically Stuck

Jay Nelson jay@REDACTED
Wed Sep 6 08:46:34 CEST 2006


Richard O'Keefe wrote:

 > And in Self an object may acquire new slots at any time without 
changing its identity.

Coming from Lisp, I always wondered why C and C++ required a storage 
location (variable) to have a type but the data or object had no sense 
of self.  The old joke was "Why, if I take my laptop to the library, 
does it become a book?"

It seemed natural to me that data have immutable identity and storage 
locations be just that: empty addresses that are references to where 
something (anything) is stored.  Why is it hard to ask an object what 
type it is, and why should its type identity change just because someone 
moves it to a different location?

I used to want to create an object system where objects are as Richard 
describes in self.  They would have identity and that's it.  There would 
be "bundles" of related functionality that could be added or removed 
from individual objects at any point of the execution or depending on 
who was accessing the object, so that things like security or 
concurrency would be trivial -- just take away all the write access 
bundles.  I even toyed with creating Aspect-Oriented Erlang which was 
process per object with meta-messages to add and remove bundles 
(aspects).  Bundles implemented methods as message handlers.  (The 
biggest trick to a new OO paradigm is to avoid the existing terminology 
so that people don't get confused with the new twist on a term that they 
already have ingrained as a particular implementation and meaning.)

Fortunately, I came to my senses and started thinking in the erlang way 
again.  I have already done my stint in the OO world and solved the 
problems there, but found the code tedious and verbose.  I am now back 
to functional (which I like more than any other type of code for most 
problems) and am trying to come up with succinct solutions to problems 
that previously seemed to be too complicated to solve clearly.

I do believe that erlang would make a good platform for experimenting 
with implementations of other languages and for doing simulations, but I 
haven't ventured into any problems in those spaces.

jay






More information about the erlang-questions mailing list