Dataflow, Controlflow: how is that solved in other systems? Maybe
xploring more and writing a paper about the three categories of systems
mentioned in /CHAIMS/Doc/Details/controldata.html
.
How compares CLAM with CLOS, Corba IDL, Ada, Scripting Language
for component composition developped at University of Bern (Oscar Nierstratz),
Component C++ graphical input, Rapide, .........
Enhancements necessary to CLAM? Examples of more complex megaprograms
in CLAM using helper modules for lists etc?
Progressive and partial abstract: why? definition? examples? benefits?
similarities to something in databases (ask other db-people) or in other
fields of CS (ask Gio, check references)? What are main issues of progressive
and partial abstract? How does it work in CHAIMS? Is CHAIMS limiting or
supporting for it? How would it look like in other languages? What is the
connection between autonomous megamodules and progressive/partial extraction?
How to write wrappers for it? Special difficulties with partial extract?....
Server availability, error-handling: Accounting of services,
problem of unavailability of services (temporarily, permanent, being unavailable
from the start of the execution of a megaprogram, becoming unavailable
during execution). How is communications failure, network partitioning,
etc. addressed?
Evolution of systems build by a CHAIMS megaprogram: how are changed
services introduced into an existing system? versioning? what happens if
the interface of a service changes, or a service disappears? notification
of megaprogrammers and end-users? configuration management? changes in
quality?
Repositoy: why not use one of the current available directory services
(see also Java classes for accessing various kinds of naming and directory
services)? These publicly available services would solve the problem of
distributing and accessing the repository. How is our repository distributed?
Updated at all client's sites immediately after changes have taken place
at a provider's side?
Other remarks of paper referees:
A more extensive review of relevant literature is necessary, because
most references are concerned either with one special field of examples
(molecular biology) or deals with basic literature. .... Other fields of
interest should be cited.
Missing:
a more detailed comparison with CORBA, especially the added value by CHAIMS
compared to the usage of Server Application Objects in CORBA
why is RMI over IIOP not enough?
comparison with Enterprise Java Beans
comparison with SWAP
"incorporating knowledge embedded in programs" is a reasonable claim for
megaprogramming, but the nature of hwo CPAM allows interaction with that
knowledge appears to be seriously limited by CPAM's simplicity.