Vat Location Service, Meta Tracking, mutable namespaces (was: [E-Lang] A more critical question (was: an example impatience policy))

[zooko@zooko.com]

[Cc:'ing e-lang, mojonation-devel and p2p-hackers, as this is a discussion about
the possibility of combining forces on this central architectural component.
For background, please see these articles or threads: [1, 2, 3, 4] and probably
a dozen more like them of which I am unaware.  --Z]


 Alan Karp wrote:
>
> I just read Zooko's metatracking page and realized that his system is very
> similar to the e-speak advertising service.  It appears that Zooko is far
> ahead in the area of scalability, data integrity, and what e-speak calls
> "community formation", the selection of places to search.  On the other
> hand, e-speak appears to be ahead in the representation of metadata, search
> mechanisms, and the security aspects.  Sounds like a natural collaboration
> if the parties involved have the bandwidth.


This topic is high on my mind right now.  I am still chewing on ideas and I'll
write about it soon.

One thing to be careful of is that "A Plan For Easy Distributed Meta-
Tracking"[5] has not been fully implemented yet.  Most significantly, current
Mojo Nation brokers don't actually keep a local, dynamically maintained list of
meta trackers to use -- instead they just use whatever metatrackers are listed
in the bootpage at `http://www.mojonation.net:25000/bootpage.txt'.

Of the other six "implementation steps" in the Plan, #1 is sort of optional (it
might work without it, and it might fail to work even *with* it, although #1
should definitely help), #2,3,4 are done, #5 isn't done (it just means: "manage
your local, dynamically maintained list of MTs so that you have an idea of which
ones you prefer to use for which purposes") and #6 is questionable in my opinion
-- it might be overkill.

Then there is the big question: will the system described in the Plan *work*?

Without #1, or some other technique like VLS'es "here are the MetaTrackers which
know me", or a hash-matching hint (this is the technique that I was dreaming
about last night), it is possible that you might be unable to find the current
contact info for a given peer, because he published to a set of MetaTrackers
none of whom you know.  (Actually, even *with* any technique that I can imagine,
you might still suffer this fate, so any argument about this is going to have to
be about probability and intuitions rather than something which can actually be
proven to work or proven not to work.)


Regards,

Zooko

[1] http://www.eros-os.org/pipermail/e-lang/2001-March/004904.html
[2] http://zgp.org/pipermail/p2p-hackers/2001-March/000030.html
[3] http://zgp.org/pipermail/p2p-hackers/2001-March/000028.html
[4] http://www.parc.xerox.com/istl/groups/iea/papers/plsearch/index.html
[5] http://cvs.sourceforge.net/cgi-bin/cvsweb.cgi/~checkout~/evil/hackerdocs/distributed_metatracking.html?cvsroot=mojonation