+1 for this change.
On 5 Apr 2011 07:45, "Martin Sustrik" <[email protected]> wrote:
> On 04/04/2011 11:32 PM, Paul Colomiets wrote:
>
>>> 2. Publish/subscribe. In this case we assume infinite feed of messages
>>> that individual subscribers consume. When shutting down, the subscriber
>>> has to cut off the feed at some point. Whether it cuts off immediately,
>>> dropping the messages on the fly or whether it waits for messages on the
>>> fly to be delivered is irrelevant. The only difference is that the point
>>> of cut off is slightly delayed in the latter case.
>>>
>> Currently if subscriber disconnected for some reason, it can connect and
>> continue where it left over (if its identity is same). But it's useless
<snip>
>
> Yuck. That's still in the codebase. I've almost forgot about that.
> Thanks for reminding me!
>
> Here's my proposal for 3.0:
>
> Disable identities for PUB/SUB pattern.
>
> Rationale:
>
> Identities can be used with PUB/SUB to improve reliability of message
> distribution. However, the messages that are on flight during the
> disconnection/failure are dropped anyway. Moreover, if failure lasts for
> long enough, HWM on publisher is ultimately reached and new messages are
> discarded. Which makes the whole identity thing in PUB/SUB more or less
> useless.
>
> What's more, identities in PUB/SUB hurt robustness of the system as they
> tie resources on the publisher to a dead subscriber. Scenario:
>
> Connect to publisher in loop using generated identities and disconnect
> immediately. In short time, publisher will run out of memory, as it has
> to store messages for those non-existent subscribers.
>
> Even if there's limit for number of concurrent subscribers, the fact
> that "session" remains open even though underlying TCP socket is closed
> means the same scenario would quickly exhaust the session limit and
> create a DoS situation for other users.
>
> Martin
>
> _______________________________________________
> zeromq-dev mailing list
> [email protected]
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to