From: Tim Bradshaw
Subject: C++ gets dynamic redefinition (but does it wrong)
Date: 
Message-ID: <ey3d861zbg0.fsf@todday.aiai.ed.ac.uk>
There is an article in the most recent `;login:' -- the Usenix
magazine -- describing a presentation at the 1998 Usenix conference of
`dynamic C++ classes'.

The main feature they want is that you can redefine classes in a
running system (presumably by dynamically loading an object file with
the new definition in).  They want this because they want to write
programs (network servers, databases ...) that can be patched at
runtime.  Funnily enough, whenever I've suggested to C++ people that
this might be a useful feature of CLOS they've been dismissive!

The thing they get wrong (I think) is what happens to existing
instances -- rather than having the equivalent of
update-instance-for-redefined-class, they allow multiple versions of a
class to exist at once.  There seems to be some support for
invalidating instances of an old version, but I don't think they can
be updated.  You also can't change the class interface at all, so
that's limiting.

The article described it as being like corba or activeX but didn't
mention CLOS of course.

It's kind of funny how these things get reinvented.

--tim

(I don't have any reference to the authors, I think one of them is at
AT&T)
From: Erik Naggum
Subject: Re: C++ gets dynamic redefinition (but does it wrong)
Date: 
Message-ID: <3121713617954595@naggum.no>
* Tim Bradshaw <ยทยทยท@aiai.ed.ac.uk>
| It's kind of funny how these things get reinvented.

  an old .signature has been reinstated for the occasion.

#:Erik
-- 
  those who do not know Lisp are doomed to reimplement it