On Thu, Dec 11, 2008 at 9:43 AM, Marco Pesenti Gritti <[EMAIL PROTECTED]> wrote: > Just fyi, I submitted a gadget fedora package for review. It's going > to require ejabberd 2.0.2. > > https://bugzilla.redhat.com/show_bug.cgi?id=475971
When you have a bit of time, I am still keen on understanding how does Gadget fit in the wider picture. With the "OMG! ejabberd's memory and roster mgmt are Out Of Control" thing mostly behind us, I want to have a clear picture of why and how Gadget fits into the picture. And at what (cpu, memory) cost, specially for the 3K users scenarios we're looking at. In other words, we are finding that ejabberd is very amenable to having its behaviour changed in interesting ways with - minimalistic Erlang plugins - poking at its internal mnesia DB via the xml-rpc plugin - using an external DB instead of Mnesia, and having external programs manipulate the DB all of these things keep us on using standard XMPP (so our client and server are more generic, and interchangeable) and from a scalability POV avoid adding additional processes (except for the DB). OTOH, I'm not against Gadget. It's just that it's a big unknown to me at a stage where -- without that much effort -- we seem to be getting ejabberd under control and playing nice. cheers, m -- [EMAIL PROTECTED] [EMAIL PROTECTED] -- School Server Architect - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff _______________________________________________ Server-devel mailing list [EMAIL PROTECTED] http://lists.laptop.org/listinfo/server-devel