[OPEN-ILS-DEV] The use of Jabber on Evergreen

[Mike Rylander]

On 9/12/07, Edward Corrado <corrado at tcnj.edu> wrote:
> Hi All,
>
> Let me preface this by saying I don't personally have an issue with
> Jabber being used as a communication tool in Evergreen, and I think the
> experience in Georgia shows it does the job, however.....
>
> On multiple occasions when talking to people who have varying levels of
> IT/programming background, but not too much experience with Evergreen or
> other ILS software, they start to wonder about the choice of Jabber to
> communicate within the software package. I hear things like:
>
>     "it is a weak link"
>     "it doesn't scale"
>     "it is a hack"
>     "why would they do that, Jabber isn't designed for use that way"
>     etc.

Wow ... quite a nice little list of strawmen.  Let's knock them down, shall we?

"it is a week link"
 *  First, the statement is ill-defined.  Weak link in what, and in
what way?  For the sake of argument, let's suppose that the person
making the statement is referring to the overall software stack.  XMPP
is a Message Passing Protocol, nothing more and nothing less.  Server
implementations can be clustered for both performance and
availability.  Just ask Google and Jabber.org.

"it doesn't scale"
 *  Again, you'd better let Google know about that.  ;)  The only
explanation I can see for why a person would say this is that they
don't know about server-to-server communication that is native and
core to XMPP, and transparently useful to Evergreen.

"it's a hack"
 *  What's a hack?  XMPP?  We'd better alert the IETF, because they've
got a pile of standards and specs surrounding the technology.  Is it
Evergreen's use of a messgaging protocol to decouple logic from the
web server that is a hack?  Better tell JBOSS and MQSeries and NetCool
Omnibus and CORBA than they should shape up and stop being hacks that
nobody should use.

"why would they do that, Jabber isn't designed for use that way"
 *  We use the moniker "Jabber" because it's what people know, just
like the use of "web services" when what you're really talking about
is RPC-over-HTTP ... XMPP /is/ in fact designed to allow exactly what
we do -- pass messages between endpoints in a reliable and efficient
manner.  I'd suggest that any nay-sayers take look at
http://www.xmpp.org/ before pronouncing XMPP as an ill-designed,
non-scalable hack.  ... but that could just be me ;)

>
> As I mentioned, I think the experience in Georgia shows that Jabber
> works for Evergreen, and can at least scale to the level of a state wide
> public library system (which is bigger than anything I am proposing here
> in NJ). However, I really don't know enough about the choice of Jabber
> to respond in any other way than "the proof is in the pudding, it works
> in Georgia." Can anyone offer me any better arguments why the use of
> Jabber isn't a problem (or, for that matter, confirm that it is)?
>

Thanks, Edward, for bringing some of the misconceptions you've heard
about XMPP and Jabber to light.  I think it's obvious that none of the
common complaints listed here really stand up to even cursory
investigation.

Is XMPP perfect?  Nope.  Just like any technology, there are
compromises to be made.  But it does what is needed, does in fact
scale to the level required (and far beyond), and stays out of the way
so we can get real work done.

-- 
Mike Rylander
Equinox Software, Inc
miker at esilibrary.com
http://esilibrary.com/

> Edward
>
>
>
>
> --
> Edward M. Corrado
> http://www.tcnj.edu/~corrado/
> Systems Librarian
> The College of New Jersey
> 403E TCNJ Library
> PO Box 7718 Ewing, NJ 08628-0718
> Tel: 609.771.3337  Fax: 609.637.5177
> Email: corrado at tcnj.edu
>
>