For authentication and vCards we're using a custom LDAP backend. And XEP-0313 Message Archive Management didn't exist when we had to implement a custom solution, especially on behalf of german archive law §146AO and §239HGB. We did put a lot individual developement in a solution I can't economically justify burying now.
MongooseIM is a fork of Ejabberd. One feature I liked about it when I had a look at it (that was while ago) that it had websockets support which Ejabberd didn't have back then.
Prosody doesn't scale horizontally / cluster [1]. It's a piece of cake to deploy it, but it's simply a different niche - you won't have a highly-available system serving hundreds of thousands of users with Prosody.
MongooseIM is a fork of ejabberd. The biggest benefit now (due to ejabberd catching up with some others) is an open-source huge XMPP compliance and integration test harness (ejabberd's open-source test suite is a joke and it's not known whether any closed-source suite exists). We pay more attention to failure scenarios in distributed setups (handle node downs, clean up transient data). Refactored code / cleaner internal interfaces improve iteration speed on customizations if you're interested in extending the system.
[1]: When I researched Prosody on this topic I found out that some clustering support was in development, but as far as I could tell it was far from usable. Honestly, I'd be grateful if somebody could shed more light on this.
8 comments
[ 2.3 ms ] story [ 33.4 ms ] threadThe original got captured by my filters.
I have some huge MongooseIM Nodes in production, unfortunatly RIAK is not an Option for us.
Prosody doesn't scale horizontally / cluster [1]. It's a piece of cake to deploy it, but it's simply a different niche - you won't have a highly-available system serving hundreds of thousands of users with Prosody.
MongooseIM is a fork of ejabberd. The biggest benefit now (due to ejabberd catching up with some others) is an open-source huge XMPP compliance and integration test harness (ejabberd's open-source test suite is a joke and it's not known whether any closed-source suite exists). We pay more attention to failure scenarios in distributed setups (handle node downs, clean up transient data). Refactored code / cleaner internal interfaces improve iteration speed on customizations if you're interested in extending the system.
[1]: When I researched Prosody on this topic I found out that some clustering support was in development, but as far as I could tell it was far from usable. Honestly, I'd be grateful if somebody could shed more light on this.