I'm working on a system that uses a secure hash (probably md5 or rmd160)
as a source pointer (16 or 20 bytes + a handfull for object header 
overhead + 4 for the slot; lets call it 40 bytes).

Even in a system of 100,000 methods (which I've yet to exceed), we're
only talking 4M of memory (keeping in mind that this is only used in
development).  At todays prices of about $1.50-$2.00 / M  this is less
than a $10 cost.  Now I know that part of the Squeak culture wants to
keep it lean (i.e. the wierd varying object headers) and have the system
runnable on a Palm Pilot and other environments where RAM is more
precious.

On the other hand, What is the power of the computer to be used for,
if not for the user?  When the user is the developer, and the developer
is me, I know which way I'll decide on "efficient" use of memory.
A lot of Smalltalk's (the code base) design decisions were made in
an environment where RAM was not plentiful.  Remember doubling the memory
of a computer is a lot easier than doubling the speed.

It would be nice if I could just add my own instance variable to
CompiledMethod in the browser, but ObjectMemory has a bunch of goop
to hide oop-type-slots amongst the byte-type-slots.

-- Mike Klein

On Wed, 12 Aug 1998, glenn wrote:

> I'd encourage this. At the cost of 1 by te per method (and if my algebra is
> correct, only 1/2 byte per method if objects are 16-bit padded and 1/4 byte
> per method if 32-bit), this could be a hook for interesting source
> experiments. Enabling interesting work for low cost seems very Squeaky to me.
> 
> glenn
> 
> At 10:49 PM 8/11/98 -0300, Alejandro F. Reimondo wrote:
> >why not to use an object in the method's literal frame ?
> > capable to retrieve the source and a hierarchy of
> > source retrievers to get source from a file, a string
> > in the image, a zip file or an HTTP server
> >
> >Ale.
> >
> >
> >
> 
>