Saving the environment! =P

Alan Grimes alangrimes at starpower.net
Wed Jan 28 06:00:45 UTC 2004


om

That subject line isn't exactly appropriate but I felt compelled to pick 
a nice cleche. ;)

Anyway, I'm continuing my study of Environment and I continue to be 
stunned by the elegance of many of the design decisions... The sheer 
audacity of the "reorganizeEverything" method knocked me off my feet! 
Ofcourse it will never work as there needs to be quite a bit of 
re-design before a sphere-like systemwide factorization is possible but 
the person who wrote that function did, indeed, have balls.

I'm still a bit confused about how environments relate to processes and 
am wondering about wheather there is any work towards making 
environments secure (by seperating public and private classes) and what 
support is provided towards recursive subsystem managment... It is 
indeed a marvel that someone actually implemented recursive 
environment... In my weak and confused mind, I would give each process a 
"current environment" which will specify which environment processes its 
class refferances... However, after reviewing the design of Environment, 
I suspect this line of thought is wrong....

Another problem with Sphere's security model is that it would require a 
means to put refferances to environments in the current environment, 
violating the normal environment heirarchy...  That mite make some sense 
if I give an example...

Lets say you had environments

   A
  /   \
B     C


and you wanted to give B the power to manipulate A or C, after a 
suitable authentication protocol, you would need to make some kind of 
link... What is the "Right Way" to do this?

Anyway,

I was looking at the interpreter again...

It would _SEEM_ that the usage of pool dictionaries in Interpreter would 
help clean up the usage of excessive class variables... Are there any 
technical reasons this can't happen?

The current Unix VM _DOES_ fork worker threads, which modules do this? Why?

Is there any evil in the current toC compiler that needs to be exorcized?

It is clearly time to make squeak P4HT compatible and <sheepish grin> 
SMP enabled. ;)

The real trick to this seems to be to designing good thread primatives 
that can front-end a wide variety of host platforms... I know a tency 
wency little bit about thread programming but I only have a cursory 
understanding of the zen that makes the VM work... =\




More information about the Squeak-dev mailing list