Why won't you just tell me?

There is no correct answer to the "How many users?" question. It depends on many things. Here are some (not all) of the possible things that could affect how many users Prosody can handle:

• Prosody configuration (including enabled modules, etc.)
• System configuration (available RAM, operating system and its configuration, etc.)
• The behaviour of your users (do they have rosters? what kind of traffic are they generating?)

For example, Prosody can handle many more users if you disable all modules, all the users have empty rosters, don't generate any traffic, and don't use encryption or compression. Obviously this scenario is not very realistic, as an XMPP server configured this way would not be very secure or very useful. Yet some other server projects have published benchmarks demonstrating their impressive scalability under just these circumstances.

We'd like to provide a realistic and honest figure, but there are so many different configurations, and every setup is different. It's impossible to just choose one number.

Why won't you publish benchmarks?

For similar reasons. Because every setup is different, benchmarks from two different setups cannot be compared. Even different benchmark tools on the same system will typically produce massively different results. Running benchmarks is a tricky operation. To be done correctly you have to be very careful to run them on a clean system, preferably not in a virtual environment or anywhere that CPU could be "stolen". You have to ensure your CPU isn't doing automatic frequency scaling, and so on.

This all makes publishing benchmarks next to useless, and harmful in fact (because people will naturally want to compare them to other projects). This would make Prosody incorrectly look either better or worse than other projects.