If areaA, areaB and areaC are all in the internal network, you can just set up a single server wildfire.company.com and they all can talk to each other.
This would work, but I would like to prevent the following:
If the Site to Site Intranet (yes they are all on the same internal WAN connected network) is down only users at HQ would be able to use the IM server. With distributed servers local communication could still take place.
I’‘d like to reduce site to site WAN traffic, so intra-office IM’'s stay in their respective LANs.
if you want to set up server for each area, you still can do wildfireA.company.com,
wildfireB.company.com and wildfireC.company.com , then you can use s2s (server to
server) option to make them talk each other.
I’'ve done this but it creates some Jabber login name distribution problems:
I am using the LDAP connectivity to pull in the account names, and groups from AD. Thus on each respective server I have:
The users need to know that user1 logs into areaA.company.com, and user2 logs into areaB.company.com rather than a flat namespace. They need to add user1@areaA.company.com to their roster, in an AD scenario, this is even more confusing b/c user1@areaB.company.com also exists, but would never be logged into.
This unto itself is not the end of the world, as it is not such a big mega-corp that people don’'t know where each other work geographically. It does create confusion for any traveling salespeople however.
It does create a problem when shared rosters based on groups from AD come into the picture though. Each respective areaX server will pull in the same list of users from AD, except it will present them all as userX@areaLOCAL.company.com. So I have a shared roster saying Bob, Joe, and Sandy are all local, however in reality maybe only Bob and Joe would log into their local servers, and Sandy works at another site.
if you want to have server in each area but you only want them to use wildfire.company.com , please vote for New Feature-- Open: Unresolved JM-191
That sort of seems to relate more to High Availability failover, but maybe the problems are the same. I thought it was my understanding that this problem could be solved with clever DNS/SRV entries. I’'m looking for the advantages of a distributed server architecture, without loosing the advantages of a flat namespace.
Hope that clarified it.