tahoe-lafs wrote: > #839: Copying directories containing caps from the > future
No one has thought through the general problem of protocol negotiation, and designed a full and complete solution for capabilities. I have been intending for some time to write up a proposal on this, but hesitate because I have little practical experience with capabilities, though some experience with the protocol update problem. As always, I recommend Microsoft's MIDL/com as a example of general purpose, future proof, protocol negotiation for binary complied objects. Microsoft has been the technology leader in this field and is worthy of imitation. Actually it was not entirely future proof, because it failed when applied to inter thread, interprocess, and remote communication, but it has been rock solid future proof in the area it was originally designed for, objects inside the address space of a single thread, allowing people to casually update shared objects without crashing previously existing code of which sharing they were unaware. We need something like Midl, but designed for message passing connections between programs running message pumps, rather for object calls, and designed for message passing connections whose messages may contain capabilities, a capability being always a capability to create another message passing connection, much as Midl was designed for objects that supported function calls that could return references to other objects, objects that could have their interface designed in the future of the time that the calling program was compiled. _______________________________________________ tahoe-dev mailing list [email protected] http://allmydata.org/cgi-bin/mailman/listinfo/tahoe-dev
