squeak printing?

Lex Spoon lex at cc.gatech.edu
Fri Feb 8 23:39:18 UTC 2002


> 
> I'll go one further -- perhaps we should be thinking about remodelling 
> the FileStream and FileDirectory objects so that they can directly 
> support these kind of abstractions, whether or not the operating system 
> does?  Perhaps something along the lines of, but more tractable than, 
> Styx in the Lucifer system?  If we imposed our own file-system model on 
> the corpus of system resources, might we be able to provide, with proper 
> refactoring, for a more robust --and more readily extensible-- system?
> 
> If we did that, we could begin with trivial "drivers" as pluggable 
> primitives, and build thereupon in Smalltalk code -- the very model that 
> worked so well for the rest of Squeak.

I guess the idea is, Squeak would have an internal naming system to
handle its global data.  Instead of just plopping things on a desktop or
in flaps, you could save them by stashing them into the image-wide
directory tree.  The filesystem of the underlying OS could be mounted in
this hierarchy somewhere.  Probably, the WWW should be mounted
underneath it as well.  :)

One concern is that we might be overdoing it.  There are only 5-10
global variables other than classes in Squeak right now.  Do we need a
hierarchical naming system for this?  Are we engaging in premature
generalization?

Anyway, what would the polymorphic methods be?  acceptObject: ? 
acceptMorph: ?   acceptText: ?  retrieveObject/Morph/Text ? 
isDirectory?  Would a path now be an array of symbols, like in the
module system?  (And, would the modules hierarchy be mounted in this
hierarchy, too?)


-Lex



More information about the Squeak-dev mailing list