[OPEN-ILS-DEV] Ejabberd configuration for OpenSRF

[Thomas Berezansky]

I don't think localhost is needed.

As for max_stanza_size, it doesn't need to be a power of two because  
it isn't a memory size or anything, it is a sanity check in ejabberd.  
It needs to be larger than the default because OpenSRF sends fairly  
large messages on a regular basis, at least with Evergreen running  
over it. At the size in question S2S messages don't need a  
significantly larger setting.

In general, the default and recommended items across ejabberd's config  
is assuming you are setting up an XMPP *chat* server, with people  
sending messages to each other using the various features of XMPP.  
OpenSRF interacts with XMPP much differently, and as such the various  
notes may not fully apply.

Thomas Berezansky
Merrimack Valley Library Consortium


Quoting "Lazar, Alexey Vladimirovich" <alexey.lazar at mnsu.edu>:

> Hi, a couple of questions about ejabberd settings.
>
> 1. For OpenSRF, is the "localhost" entry required in the hosts directive?
>
> {hosts, ["localhost", "private.localhost", "public.localhost"]}
>
> Or can it be safely omitted with just the private.localhost and  
> public.localhost hosts listed?
>
> 2. Max_stanza_size.
>
> Ejabberd documentation says:
> {max_stanza_size, Size}
> This option specifies an approximate maximum size in bytes of XML  
> stanzas. Approximate, because it is calculated with the precision of  
> one block of read data. For example {max_stanza_size, 65536}. The  
> default value is infinity. Recommended values are 65536 for c2s  
> connections and 131072 for s2s connections. s2s max stanza size must  
> always much higher than c2s limit. Change this value with extreme  
> care as it can cause unwanted disconnect if set too low.
>
> OpenSRF instructions say:
> Change all max_stanza_size values to 2000000.
>
> The OpenSRF-recommended value is the same for c2s and s2s  
> connections, which differs from what the ejabber documentation  
> suggests. I am curious about a) what requires these values to change  
> for OpenSRF in the first place, b) why the OpenSRF-recommended value  
> for c2s and s2s is the same, and 3) why a power of 2 is not used for  
> it, e.g., 2097152?
>
> Any "science" behind these and other OpenSRF-recommended settings,  
> or just safe guesses?
>
> Thanks.
>
> Alexey Lazar
> PALS
> Information System Developer and Integrator
> 507-389-2907
> http://www.mnpals.org/
>
>