[jdev] [JEP-0114] Jabber Component Protocol question

[Vinod Panicker]

On 3/6/06, Peter Millard <pgmillard at gmail.com> wrote:
> As far as I know, all servers which use the component protocol, do so
> over ports OTHER than 5222. Whether you listen on multiple ports of a
> single port is an implementation decision.

It means that server implementors found it convenient not to mix
acceptors for clients and components.  I'd say that it would be better
to "officially" recommend a standard port for components.

> It would be interesting to have SRV records for this, but normally,
> since components are "trusted", they know (via config, or cmd line
> switch) the port to use.

But SRV records have other benefits - keeping the components loosely
coupled also providing scalability and fault tolerance due to SRV
features.

> On 3/3/06, Vinod Panicker <vinod.p at gmail.com> wrote:
> > Hi,
> >
> > I was just getting around to implementing JCP into the server, and
> > realized that the server would need to treat a "component" different
> > from a "user".  The JCP spec also defines a new jabber:component
> > namespace under which all the stanzas would be sent.
> >
> > IMO, it would be great for server implementors to have component
> > connections being accepted on a different port instead of the default
> > 5222.  Also, if a new SRV record could be defined of type
> > xmpp-component, components can look it up and establish connections to
> > servers accordingly.
> >
> > We are already doing it for the jabber:client and jabber:server stream
> > namespaces... why not for the jabber:component namespace?  This way,
> > the concerns can be separately and elegantly addressed.
> >
> > WDYAT?

Regards,
Vinod.