Re: [paho-dev] asynchronous MQTT protocol parser

[Andy Piper <andypiperuk@xxxxxxxxx>]

On Mon, Jul 8, 2013 at 1:39 PM, Julien Vermillard <jvermillar@xxxxxxxxxxxxxxxxxx> wrote:

On Mon, Jul 08, 2013 at 03:58:56AM -0700, Andy Piper wrote:
> Just curious whether you looked at the async client API that IBM contributed, which is in the develop branch of Eclipse Git at the moment.

The API is asynchronous but the inner socket code and parser is synchronous which won't scale much more than the current implementation.

I had not dug deeply into it, but I did realise that the async API is built on the sync one...

 

> I believe that there is also a bug open around moving to NIO at some stage but I may not be remembering correctly.

Tried to find one without success :(

I can't find it now either - not sure whether Bugzilla allows searches for 3 character words actually. I know we had a discussion about NIO at some point either on the mailing list or on Bugzilla and I think the IBM folks were uncertain at the time of the benefit, as well as being concerned about the level of Java it would prereq.

 

> Is the idea of posting this here that you are looking to contribute your implementation in the future?

Yes, I could contribute here if there is some interest.

I think, assuming that it was "in line" with the existing Java client rather than being "yet another client", that would be very interesting. Greater scalability and performance would benefit everyone.

I'll let others who have worked on the Java client chip in with any thoughts on this.

Andy

--
Andy Piper | Farnborough, Hampshire (UK)
blog: http://andypiper.co.uk   |   skype: andypiperuk
twitter: @andypiper  |  images: http://www.flickr.com/photos/andypiper