Eclipse Community Forums
Forum Search:

Search      Help    Register    Login    Home
Home » Newcomers » Proposals » Mosquitto proposal
Mosquitto proposal [message #1117880] Thu, 26 September 2013 18:29 Go to next message
Roger Light is currently offline Roger Light
Messages: 3
Registered: September 2013
Junior Member
The Mosquitto project provides a small server implementation of the MQTT and MQTT-SN protocols.

MQTT is a protocol designed for machine-to-machine (M2M) communication that runs over TCP/IP. Its primary goals are:

  • to avoid polling of sensors, allowing data to be sent to interested parties the moment it is ready
  • to be lightweight, so that it can be used on very low bandwidth connections and with limited resources

MQTT-SN is a protocol derived from MQTT, designed for connectionless underlying network transports such as UDP.

http://www.eclipse.org/proposals/technology.mosquitto/

Re: Mosquitto proposal [message #1119755 is a reply to message #1117880] Sat, 28 September 2013 17:39 Go to previous messageGo to next message
Nicholas Humfrey is currently offline Nicholas Humfrey
Messages: 1
Registered: September 2013
Junior Member
Hello,

This is an excellent proposal.

Roger has done a very good job of developing and maintaining Mosquitto but I think it would really benefit to become part of Eclipse. Having MQTT-SN support merged in would also be fantastic.

I am also far more likely to made contributions, with the source code in Git, rather than Mercurial Confused

nick.
Re: Mosquitto proposal [message #1125675 is a reply to message #1117880] Fri, 04 October 2013 15:33 Go to previous messageGo to next message
Scott Lewis is currently offline Scott Lewis
Messages: 970
Registered: July 2009
Senior Member
Greetings,

I've read over the Mosquitto proposal. ECF (Eclipse Communication Framework) is interested in being a consumer of this work, as we will be able to use an MQTT broker/server implementation to add MQTT broker functionality to OSGi runtimes (OSGi web servers, other servers, Eclipse, etc). ECF will also be able to expose standard OSGi remote services via an MQTT transport. We/ECF project can be added to the interested parties for this proposal, if desired.

The current proposal states that the existing Mosquitto codebase is written in C. Are there project plans to expose this API in Java...as a bundle/plugin? If we (or nearly any other Eclipse project) wants to consume/use this work, it will ultimately be necessary for the API to be exposed to other bundles/plugins...which currently implies a Java API. Is this work...expose MQTT API as OSGi bundle...part of the plan for Mosquitto?

FWIW, if resources are needed to do such an effort I would personally be willing to contribute some...in order to make Mosquitto more consumable by EF projects.

Re: Mosquitto proposal [message #1125799 is a reply to message #1125675] Fri, 04 October 2013 18:26 Go to previous messageGo to next message
Roger Light is currently offline Roger Light
Messages: 3
Registered: September 2013
Junior Member
Hi Scott,

It would be good to make it easier for other Eclipse projects make use of our work. What you describe with respect to Java isn't something that Ian and I have discussed however.

Mosquitto as it stands is really intended to be light weight - I get quite a bit of correspondence from people running it on platforms where java may not be appropriate such as embedded openwrt boards.

Personally I have no intention to rewrite in Java, but would be happy to receive help on bundling if it is appropriate - that's not something I have experience with. I don't see a problem with a Java version being committed either, although it might go against the proposal as it is written at the moment (would that be a problem?)

I'm a bit of a beginner when it comes to the Eclipse Foundation, but I think that some of the differences between the mosquitto proposal and others is that having had a skim of existing proposals they seem to mostly have half a dozen committers all working for the same company. I'm just a bloke that set out to write some code that then happened to get a bit popular and so I'm trying to make it more accessible to other developers and users.
Re: Mosquitto proposal [message #1128212 is a reply to message #1125675] Mon, 07 October 2013 09:05 Go to previous messageGo to next message
Ian Craggs is currently offline Ian Craggs
Messages: 14
Registered: March 2013
Junior Member
Hi Scott,

the original reason for writing RSMB in C was that we had environments that

a) were too small to run a JVM, or

b) for which no JVM was available.

Reason b) is less likely these days, with the open source JVMs, but reason a) still holds.

I think there is room for both a C and a Java implementation of the function in RSMB/Mosquitto, each fulfilling slightly different criteria. C for the smallest implementation, and Java for really nice integration into the Eclipse IDEs and OSGi stacks.

From an MQTT client point of view, of course, it doesn't matter. Any MQTT client in any language can connect to any MQTT server, assuming they implement the protocol correctly.

For Eclipse, I think the question is whether a Java implementation, if it turns up, should go into Mosquitto or a separate project. Currently I have scoped the proposal to describe a C implementation. A Java implementation would have a different scope, and therefore should be in a different project.

If there were strong enough feelings that the Mosquitto project should be a placeholder for MQTT servers of all types, then the proposal would need to be altered to reflect that scope.
Re: Mosquitto proposal [message #1130633 is a reply to message #1117880] Wed, 09 October 2013 15:37 Go to previous messageGo to next message
Nicholas O'Leary is currently offline Nicholas O'Leary
Messages: 4
Registered: November 2012
Junior Member
It is great to see both Mosquitto and RSMB get to this point; a worthwhile addition to the Eclipse M2M ecosystem.

On the java question, I'm all for a OSGi-based MQTT-only broker implementation. Java-bindings to the Mosquitto broker would be an interesting stop-gap measure, but will only take us so far. But let's not distract ourselves with the next task before the current task is done.

From a project organisation point of view, I'd be inclined to have Mosquitto be the home of MQTT Brokers as Paho is the home of MQTT clients. This may not fit with the current proposal, but I'm sure its something that can be accommodated either now, or at a later date when we have firmer plans around other brokers.



Re: Mosquitto proposal [message #1130754 is a reply to message #1130633] Wed, 09 October 2013 18:40 Go to previous message
Ian Craggs is currently offline Ian Craggs
Messages: 14
Registered: March 2013
Junior Member
Thanks, Nick and Nick (!) for your comments.

It seems like a reasonable approach to leave the decision of how to incorporate another MQTT broker into Eclipse to the point where one is actually contributed.
Previous Topic:New project proposal: Kura
Next Topic:3P - PolarSys Packaging Project proposal
Goto Forum:
  


Current Time: Wed Aug 27 11:23:37 EDT 2014

Powered by FUDForum. Page generated in 0.01732 seconds