aether
aether copied to clipboard
dtqec
Distributed system emulation in Common Lisp
Almost everywhere in the blossom algorithm, we are careful not to replicate references to objects (especially among different processes!) and instead make copies of the value to share and edit....
When a command is part of a single script, that should be made clear by a consistent naming convention. We often do this by hand, but it would be better...
The communication manager is not especially good at tracking lost / leaked objects. Primarily, these include unanswered (and unanswerable) messages, as well as channels that are allocated and discarded without...
A lot of simulation processing power is spent on `PROCESS`es which are busy-waiting for the receipt for some message. This could be alleviated by providing a callback to the `COURIER`...
Currently, `ADDRESS`es carry both the fields needed to identify which mailbox it points to and a `SECRET` field that indicates ownership of the mailbox. It is currently inconvenient to make...
It is unrealistic from the perspective of hardware to keep a full list of commands-to-be-executed for a running process. Rather, processors tend to carry a fixed set of instructions, a...
Many of the command definitions start with an `(unless aborting? …)` wrapper around their body. It would be preferable, I think, for this clause to be part of the procedure...
Messages are not currently named consistently, stored in consistent places in files, … . Their organization has some of the same flavor as the organization of commands making up a...
In order to ensure program correctness, we need to be very consistent about not letting the author of software-to-be-emulated communicate between objects by any means other than message-passing. A good...
Currently, we use CLOS's ability to dispatch on keyword `EQL`ity to link items in a `PROCESS`'s command stack with their executable definitions. This is deficient for compile-time checking: CLOS can't...