Partitioning the image (was Re: Shrinking sucks!)
Cees de Groot
cg at cdegroot.com
Tue Feb 15 15:44:49 UTC 2005
On Tue, 15 Feb 2005 08:40:19 +0200, Brent Pinkney
<brent.pinkney at aircom.co.za> wrote:
> This just seems lacking from the discussion.
>
Not completely - I've made the point but not in a separate posting, and I
threw in lots of stuff to make sure that it wouldn't stand out ;).
But, yes, the first discussion should be - do we need namespaces? Why do
we need namespaces? A lot of the discussion has an implicit 'yes' to the
first question and probably the second question has been answered a long,
long time ago but I always find it good to keep asking these questions
over and over again, because circumstances, insights, people change.
For example, coming from <uh oh...> Java, I though namespaces quite
logical. Python, Perl - the other languages I used at the time, also have
it. Then I came to Squeak and found it lacking in this area. It hurt. Then
I moved on to VW, and made copious use of its namespace features. When
coming back to Squeak, boy was I glad that I was relieved of the burden of
having to think about namespaces! Prefixes rock compared to VW's
namespaces.
Now, that doesn't mean to say that every implementation of namespaces is
bad, but it did lead me to re-ask the question. Do we need namespaces?
And if I then consider that:
a) you only need to have one Socket, and if it doesn't work, fix it;
b) I actually don't mind prefixes at all;
c) the way forward to really really large systems is probably partitioning
in processes/islands/subimages/vats/networks/... where cross-partition
communication should ideally be based on as little naming as possible and
more on abstract things like UUID's (or in a completely orthogonal naming
space like URL's - just an example, I'm *not* proposing to replace class
names with URL's ;));
d) conflicting names can be talked about and resolved on a human level. I
usually prefer human solutions over technological solutions, and
Smalltalk's philosophy seems to be with me there;
my current position is that we can do very well without them.
More information about the Squeak-dev
mailing list
|