[zeromq-dev] Persistent queues

Mike Owens mikeowens at gmail.com
Tue Jun 28 22:54:56 CEST 2011

I was just about to ask about this very thing: durability and identity
should be two different things, and it seems that at present they are tied
together. I could care less about durability, and agree that it should be
built on top of the core, but I think explicit identity is nice to have.
It's easier for me to use explicit naming for peers rather than having to
pass a message through a router, extract the default assigned UUID and send
it back just to figure it out.

So are you saying you are getting rid of durability AND identity, or just
durability? It would be nice to keep identity, but if it is scrapped, it
would be nice to have an API call to get the default generated UUID.
Suggestion would be to have zmq_getsockopt(socket, ZMQ_IDENTITY, ...) return
it. It does not seem to provide this for default generated UUIDs.

-- Mike

On Tue, Jun 28, 2011 at 3:34 PM, Pieter Hintjens <ph at imatix.com> wrote:

> Hi Garrett,
> The comment was about phasing out durable sockets. The socket queues
> are not really message queues, it's better to think of them as
> analogous to network buffers. Allowing these to be durable seemed to
> be a good idea but has turned out to be a major pain for the actual
> design of 0MQ. It's the single major cause of nastiness in the core
> library.
> Given that it's really expensive to support (and this nastiness stops
> us from cleaning up the library to make it easier to do other stuff),
> we have to ask, is this use case really valid?
> Persistent queues - properly - should be built on top of 0MQ, as shown
> by models like Majordomo. You need to qualify "persistence" properly.
> It might mean saving to disk, it might mean transactions,
> acknowledgements.
> For durable pubsub, we've got the Clone pattern (see Ch5 of the
> Guide). This is far more useful than durable sockets.
> The only use case I've seen (recently) for explicit identities is
> router-to-router connections, which we need in some p2p architectures.
> However those don't require durable sockets, just explicit identities
> that allow unsolicited sends. So it's a subset of today's explicit
> identity / durable socket functionality.
> tl;dr: Martin Sustrik really wants to remove explicit identities from
> 3.0, and it seems a fair trade-off for other more useful features.
> -Pieter
> On Tue, Jun 28, 2011 at 2:19 PM, Garrett Smith <g at rre.tt> wrote:
> > I recently saw a comment from Pieter about phasing out persistent queues.
> >
> > Assuming I got that right, could someone elaborate a bit, or point me
> > to a thread?
> >
> > Garrett
> > _______________________________________________
> > zeromq-dev mailing list
> > zeromq-dev at lists.zeromq.org
> > http://lists.zeromq.org/mailman/listinfo/zeromq-dev
> >
> >
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20110628/d46b69ad/attachment.htm>

More information about the zeromq-dev mailing list