Re: [hiprg] draft-ahrenholz-hiprg-dht-05

[Ken Rimey <[e-mail]>]

On Mon, Nov 2, 2009 at 7:28 PM, Ahrenholz, Jeffrey M
<[e-mail]> wrote:
> what is the ideal interface?
> ...that would be a great discussion for this list.

> ...we may need to include something like the SEQ TLV in the HDRR,
> so that each record has an update id; then clients may retrieve all
> available values and use the most recent one

A better interface (I won't dare to say ideal) would handle this on the
server side, so the client only sees the most recent update.

Another desirable thing, as made clear in the draft, would be for the
server to do the checking of signatures.

The following is not a sales pitch.  :-)

OpenLookup actually implements an interface like this for a second
datastore, separate from the OpenDHT-compatible store.  The data
model is a single-valued mapping from principal-label pairs to values
(where the label is analogous to the OpenDHT key).

The interface is based on HTTP, but not XML-RPC.

Instead of a sequence number, it simply uses the expiration time
to determine precedence between updates.

For HIP, unfortunately, the OpenLookup code would have to be
hacked to name principals using just the middle 100 bits of the
SHA-1 hash.

The interface also allows for queries, where the principal is a
wildcard.  This is useful for use cases like storing a name-to-HIT
mapping.  (I have doubts about such multivalued mappings being
viable in the real world, however.)

Ken