dana@gatech.CSNET (Dana Eckart) (03/11/86)
I have been thinking about the concepts upon which smalltalk is designed and something occured to me....why not (instead of having a file system) have an "environment" system? The idea is that what would be called directories in unix would be called environments. The idea of files would be replaced by the class definitions themselves (data objects being particular instantiations of these classes). Once in an environment, anything in that environment (or "higher" up in the directory structure) could be accessed. Further, the directory structure of the environments can be DAGs (and not necessarily trees). [NOTE: I realize that the files currently used in smalltalk are instances of classes, what I'm considering is that the environment would behave (in part at least) as if upon entering a directory in a file system, all of the "code" would be "loaded" automatically.] As an example, suppose that you had a directory which defined a "virtual vax" and were curious what would happen to a network of these machines. Since there are many different kinds of networks, you might create a sub-directory for each different type of network....the nodes of the network (the vaxen) would be inherited from the parent directory. If my memory (foggy as it is) is correct, this idea resembles the package notion used by Symbolics for Zeta-Lisp (I don't know about LMI), although I think the package hierarchy had to be a tree (?). If environment systems are used instead of file systems...this seems to be more in line with the idea of smalltalk being a simulation language (and makes the system correspond more closely to the language). How do other people feel about this? Has this ever been proposed (has it actually been done?)? (Is this even presented in a coherent manner?) Anxious to hear from the rest of the net.... --dana -- --Dana Eckart Georgia Insitute of Technology, Atlanta Georgia, 30332 ...!{akgua,allegra,amd,hplabs,ihnp4,seismo,ut-ngp}!gatech!gitpyr!dana