robert.wills at dicta.org.uk
Sun Aug 8 04:38:28 MSD 2004
On Sat, 2004-08-07 at 18:51, Mickael Remond wrote:
> Just a question. What were your motivations for dropping jabberd ? It is interesting to know to help convince other people.
This wasn't addressed to me directly, but since I'm in a similar boat &
thinking of migrating from jabberd2 to ejabberd, I thought I'd reply.
Reasons for choosing ejabberd:
1. erlang: I'm a newbie to erlang but it strikes me that it's the right
language for implementing a jabber server:
- safer than c (jabberd & jabberd2 have had their share of memory
leaks) -- my understanding is that this won't happen with erlang
- lightweight threads: jabberd & jabberd2 use operating system threads
which means that with lots of users you can run into problems with
running out of file descriptors -- of course, you can increase the file
descriptor limit, but IMHO lightweight threads make more sense
- concurrency-oriented: this allows you to produce a better jabber
cluster where users can register on any node as opposed to jabberd where
you have to allocate users to individual nodes
- hackability-- as I said I'm new to erlang but even so I think it's
easier to understand the source code and see how additional features
could be added than with jabberd (I added the ability for an
administrator to add new users through the configuration node using
tkabber & it seemed to work (I'm happy to submit this patch if anyone's
interested btw)) Also the fact that erlang is safer again makes this
less intimidating than with jabberd.
2. Configuration interface: The ability to reliably see who's online,
etc is a big plus for convincing managers.
Probably the main point contra ejabberd is that it's a smaller project
than jabberd2 with fewer people on the mailing list, but activity this
list seems to have picked up since 0.7...
More information about the ejabberd