[Standards-JIG] pubsub: cache-last-item

Peter Millard pgmillard at gmail.com
Wed Feb 1 17:57:36 UTC 2006


The whole point of having persistence be an option is to reduce the
memory/db footprint of a server implementation that may need to
support billions of nodes (for example). Requiring the caching of last
items is really bad IMO and conflicts with the rest of the spec which
allows you to build event-only systems or systems which are completely
transient (don't persist any items).

They should remain SHOULDs in -60.

pgm.


On 1/30/06, Peter Saint-Andre <stpeter at jabber.org> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> I want to make sure that the proposed "cache-last-item" functionality is
> clearly understood before we move forward with revisions to JEP-0060. My
> understanding from the list discussion last fall was this:
>
> 1. Even if a node or service is not configured to support item
> persistence, it SHOULD cache the last item published to each node.
>
>    Question: should this be MUST?
>
> 2. When a subscription is accepted/approved, the service SHOULD send the
> last published item to the new subscriber.
>
>    Question: should this be MUST?
>
> Is this consistent with our discussion last fall?
>
> Note that this obviates the need for a "subscribe-and-get-last-item"
> feature.
>
> Peter



More information about the Standards mailing list