History: Why Openfire
Preview of version: 61
Openfire has been selected as one of the Software Components for WikiSuite. Here are the general component criteria.
Why Openfire instead of the many options in this space? To properly explain this one, it's important to first distinguish the various "real time" use cases. All use cases need chat, most need audio and screensharing. However, there are some "key distinctive features" which make some tools great at one use case, but poor for another. Try it out now at http://wikisuite.chat/
Realtime collaboration use cases chart
| Type | Predominant mode | Key distinctive features | Typical app |
| Ongoing Team collaboration on projects | 1 to 1, many to many or emergent | Presence, and can escalate to audio / video / screensharing as needed | Skype |
| Meetings / conference calls | Many to many | Meeting notes (meeting agenda, and live collaborative note taking for decisions) | Etherpad + phone call, or Skype |
| Webinars / Scheduled Course | 1 to many | Presentation and whiteboard | BigBlueButton |
| Community presence and support | many to many | web interface and desktop/mobile clients | IRC |
| Help desk for team members (Remote Assist) | 1 to 1, but can be transferred | Share screen and remote control. Easier to install software on their computer. Team member must give permission to take control of computer (ex: for 30 minutes) | TeamViewer |
| Help desk for customers | 1 to 1, but can be transferred | To route request to someone who is available. Canned responses. Difficult to install software on their computer. | Openfire Fastpath |
| Remote Management | 1 to no one or 1 to 1 | Remote login and management, even unattended | VNC / Guacamole / RDP |
We picked Openfire for the following reasons:
- XMPP support, and thus presence (using standards)
- WebRTC support (via inclusion of Jitsi Meet)
- Great admin panel
- Vast feature set
- The community
- Generally satisfies the usual component criteria
We ultimately want WikiSuite to be awesome at covering all the use cases above, and we do so by mostly combining:
- Server-side: Openfire + Converse (XMPP) with Jitsi Meet (WebRTC)
- Desktop client: Pàdé, a Chrome extension with tons of features. See Pàdé Presentation
Why XMPP is important
https://xmpp.org/about/technology-overview.html
https://xmpp.org/about/myths.html
Why not BigBlueButton
Tiki Wiki CMS Groupware has built-in (but optional) integration with BigBlueButton since Tiki 5 (2010), and the two communities worked closely together for a tight integration and thus, for WikiSuite, this would have been the simple option.
While Openfire Meetings and BigBlueButton broadly share the same feature set (videoconferencing, screensharing, etc), there are fundamental difference. BigBlueButton is a distance education tool.
- So the focus is one to many.
- No presence feature (it's for a scheduled class, and not ad hoc collaboration)
- No XMPP support (ex.: Federation)
- Still in 2017, the Flash version is the main one, and the HTML5 version is not ready for prime time
For WikiSuite, it's critical to also cover all the other use cases above.
The leaders of both projects (Fred Dixon and Marc Laporte) met 2 or 3 times over a few years to try to find a way to make it work. But ultimately, the basic DNA / drive / philosophy / focus which made BigBlueButton successful for distance education lead to some design choices (ex.: no XMPP) which are hard to change afterwards.
Why not Etherpad or Hackpad
- Too few features
- No XMPP support
https://github.com/dropbox/hackpad
Why not Jitsi Meet (standalone)
- Jitsi Meet is part of the solution (WebRTC), but alone is not sufficient to cover the desired use cases, which is why WikiSuite uses Jitsi Meet as part of Openfire. This is similar to how Jitsi Meet is part of Atlassian HipChat
Why not Apache OpenMeetings
Apache OpenMeetings is an interesting option with a diversity of paid support options and quite a few features, however, the focus is more about scheduled meetings or classes than ongoing collaboration. For example, XMPP is not supported.
Why not WebHuddle
WebHuddle is inactive: https://sourceforge.net/projects/webhuddle/ and requires Java applets which are restricted by browers nowadays: WebHuddle "works in most web browsers enabled with Java."
Why not Spreed
This was the situation:
- Spreed WebRTC implements a WebRTC audio/video call and conferencing server and web client.
- No XMPP
- low activity level
But then NextCloud reactivated the project: https://github.com/nextcloud/spreed/ (well done!)
There is discussion on XMPP: https://github.com/nextcloud/spreed/issues/826 and link to a client: https://github.com/nextcloud/jsxc.nextcloud but where is the XMPP server?
Why not Hubl.in
Hubl.in is part of OpenPaaS, and is a newer option. Social networking + videoconferencing + realtime collaborative editor + others.
This is WebRTC (which is great) but it doesn't handle XMPP.
https://github.com/linagora/hublin
Why not Tox
Tox is interesting
- But no XMPP
- Serverless aspect of Tox doesn't have huge value for us because WikiSuite is a server, and we have Syncthing for P2P file sync.
Why not Retroshare
Retroshare is very interesting
- But no XMPP
- Serverless aspect of Retroshare doesn't have huge value for us because WikiSuite is a server, and we have Syncthing for P2P file sync.
- Retroshare is more focused on disseminating files, than on collaborating on files
Why not Zulip
Not based on XMPP
Lots of chat features but what about videoconference?
Why not Otalk
Why not Tigase
Tigase is an XMPP server and it could have been the base
Tigase and Openfire are both written in Java.
License: Tigase is AGPL, Openfire is Apache
Openfire has more users and contributors
Why not ejabberd
ejabberd is an XMPP server and it could have been the base.
ejabberd has two editions:
- Community Server (eCS) (Free/Libre/Open Source)
- Business Edition (eBE) (not Free/Libre/Open Source)
As of 2018-10, many features you'd expect for a typical project are only available in the Business Edition (eBE) (not Free/Libre/Open Source). This open core model is a disincentive for community contributions. In contrast, these features are Free/Libre/Open Source in Openfire. And if anything is missing in Openfire, we can contribute and everyone in the community will react favourably.
See also:
- https://feedback.process-one.net/support/solutions/folders/6000076557
- https://www.process-one.net/en/ejabberd/protocols/
- https://github.com/processone/ejabberd
ejabberd is written in Erlang: not a deal breaker, but less known than Java, used by Openfire, Elasticsearch and Kibana.
Why not Prosody
- Prosody is an XMPP server. This could have served as the foundation.
Prosody is very good for a multi-tenant use case.
Prosody is written in Lua (not a deal breaker, but less known than Java)
Prosody is lacking a web admin panel
There is no WebRTC support or plugin
Why not Mattermost
- Although there is a XMPP bridge, this is not a real XMPP server
Why not Matrix.org
- Interesting but not XMPP
Why not Rocket.chat
- not XMPP
Why not Let's Chat
- not XMPP
http://sdelements.github.io/lets-chat/
Why not CandyJS
CandyJS is interesting. And we experimented with it to be the XMPP client and contributed the conversion to Bootstrap for responsive design.
But:
- it's mostly designed for team chat (vs 1 on 1 chat)
- Converse has both a pop up chat and a full page interface
- activity level is low
https://candy-chat.github.io/candy/