Pekka Savola wrote: > (I'm using Tony's message only as a springboard..) > > On Tue, 9 Aug 2005, Tony Li wrote: > >> Does it? As Joe Touch pointed out, even with a large amount of physical >> memory, there are distinct advantages to having virtual memory. True, >> you may give up demand paging, but you may still with to have an >> independent uniform, and predictable address space for programming >> simplicity. These same needs drive us to create VPNs regardless of the >> size of the address space. > > > This may or may not be obvious, but I don't think anyone said it > explicitly... > > I think what PekkaN has looking for is the separation of using addresses > as a "virtualization layer" to using [something else]. If we assume > that this kind of virtualization is needed by the architecture, we'll > either need to stick to doing the virtualization using addresses, or > using something else (which may not exist yet, leading to the issue of > something missing in the architecture). > > What could those other means be? > > Well, for example, > - more flexible lookup mechanisms than single-faced DNS (the results of > the lookup might depend on which virtualizations each node "belongs to")
That's a part of virtual networking; as has been noted, you need more than just virtual addresses. You also need a separate DNS that takes the overlay as context to the node lookup, i.e., "gethostbyname(nodename, overlayname)". > - easier security matching (in case you use a specific range of IP > addresses to imply trustworthiness), e.g., using more advanced lookup > methods > > These are the issues that will come up more strongly with id/loc split. Even an id/loc split still ends up benefitting from virtualization. Joe
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Int-area mailing list [email protected] https://www1.ietf.org/mailman/listinfo/int-area
