[zeromq-dev] RFC 29 - PUBSUB clarification

Pieter Hintjens ph at imatix.com
Tue Sep 9 12:05:30 CEST 2014


As the HWM is 1,000 by default, you won't in practice drop
subscription messages. I think these semantics haven't been reviewed
very deeply, and could be improved (e.g. infinite HWM on XSUB outgoing
queues would make more sense IMO).

On Tue, Sep 9, 2014 at 10:19 AM, Merijn Verstraaten
<merijn at inconsistent.nl> wrote:
> Also, it is very unclear what the following sequence of events should result in:
>
> 1. create socket
> 2. connect to host A
> 3. subscribe to FOO
> 4. connect to host B
> 5. subscribe to FOO
>
> Does this result in 2 FOO subscriptions at A and 1 FOO subscription at B? What if a subscription message is silently dropped? How can I tell? If I unsubscribe from FOO am I supposed to only have 1 FOO subscription at A left? Again, what if an unsubscribe is dropped?
>
> Cheers,
> Merijn
>
> On 09 Sep 2014, at 00:44 , Merijn Verstraaten <merijn at inconsistent.nl> wrote:
>> Ola!
>>
>> I was going over the PUBSUB RFC and the behaviour for XSUB specifies:
>>
>> SHALL silently drop the message if the outgoing queue for a publisher is full.
>>
>> Doesn’t this mean unsubscribe messages can silently be dropped? How can a sender ever be certain they successfully unsubscribed from a subscription? Especially given the fact that subscriptions are not idempotent, i.e. subscribing twice and having one of your two unsubscribe’s dropped will keep you subscribed?
>>
>> Cheers,
>> Merijn
>> _______________________________________________
>> 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
>



More information about the zeromq-dev mailing list