senn at maya.com
Fri May 12 18:17:02 CEST 2000
Andrew Maizels <andrew at one.net.au> writes:
> pohanl at my-deja.com wrote:
> > This new language would have components with no constraints on values
> > passed to it.
> > as for the values, instead of being dedicated to a method in a
> > component/class/function, it would be placed on a bloodstream/"bus".
> > cells/components would grab it when they have appropriate functions
> > that can use it.
> You seem to be talking about Linda. Do a web search on "tuple space".
Ah -- the thing about Linda/Tspaces is that the transaction model is
limited in a way that is unlike the physical universe -- there is no
way to send a message to another computational entity w/o incurring
the cost of opening a persistent memory segment somewhere... If you
are writing production-systems this might seem like a good idea, but
"natural" information systems tend to operate in more a data/message
flow-like manner. IMHO - persistence should be layered on messaging
and not vice-versa. [I just re-read this paragraph -- I'm not sure
I'm communicating very well....]
pohanl's comments on the history of systems and the stack are
interesting: The heavy architectural reliance on the "process/thread
function argument/return stack" is a holdover from the days of
non-componentized systems -- it is a fine way to maintain state while
traversing a hierarchical structure (e.g. evaluting a functional
expression), but as a way to maintain running "component system" state
it has severe shortcomings....
More information about the Python-list