>I really don't think we should try to fill that gap - since nobody >had scalability issues with citadel so far. Just for failover drdb is >perfect. There was a time when we wanted to try to be an "enterprise" mail/groupware system. That gave rise to the Global Address Book and the idea that you could spread a single Internet domain across a large group of Citadel servers, and there is still all sorts of bizarre message routing code to make that happen.
Guess what? Everyone who builds those type of systems, is using Exchange. Too many servers, too much complexity, too much licensing cost, too many overpaid MCSE's arguing with each other about the configuration. And the people in charge of buying those systems assume that for it to be worthwhile it has to be expensive. F**k those people. Citadel is for smart people. Citadel is for people who know that a single, properly configured back end server can handle thousands of seats. We run fast and efficiently because we don't run on Windows. When we need to scale out, we move the user interface engines onto separate servers, we add front end relay servers with their own spam and virus handling, etc. So I'm ok with the back end being a SQL server, but it's not going to be a SQL back end, if you know what I mean. Just indexes and blobs. Because we can't really do a true schema without dropping support for the existing NoSQL engine, and that's something we really can't do. First things first though: get everything ino the db. config comes first.