Vat Location Service
Mark S. Miller
Mon, 11 Jan 1999 16:52:08 -0800
At 04:12 PM 1/11/99 , Eric Messick wrote:
>This is for the registering Vat. It has a full SturdyRef to the VLS
>that it uses when registering. One reason for this could be to keep
>the number of registrants at a particular VLS down by not advertising
>the SturdyRef widely.
[?] Uh Oh.
This sounds like a good idea, but I think it might have been dropped moving
from OriginalE to E. Bill?
>Better would be to keep a list of what IP addresses could register.
If you're trying to limit some authority, like the authority to register,
why would anything be better than a capability??
>In any case, I've always thought that the protocol should be one that
>other services could use without the E connection setup protocol.
[#] Another good issue to settle on Thursday. Some tradeoffs:
On one hand, we've got something we think works (and descends from one you
wrote that worked wonderfully well). A short term winner.
On the other hand, many besides E would find such a routing fabric useful,
and we'd benefit tremendously if it were widely adopted. Dependencies on
the object/capability/proxy-comm level probably inhibit wide scale
adoption, as it will be work to duplicate the logic of this outside Java.
Dependencies on data-comm seem fine, as the two are naturally bundled
together, and are independent of Java. We'll probably only see VLSs grow
into a robust packet mixing network (to evade traffic analysis) if we get
wide adoption. Ultimately, we need security from traffic analysis to
create jurisdiction-free commerce. A long term winner.
Given an adequate short-term winner, why not deploy it? Often the answer
is fear of lock-in preventing one from upgrading to one's long-term
preference. I don't see a big lock-in danger here, since the inherent
fault-tolerant redundancy of the abstract VLS protocol should allow
multiple concrete VLS protocols to coexist fine in a transition.
OTOH, Bill's observation about VLSs as CGIs at ISPs may be the more
important issue, and it also argues for the long term winner.