On Wed, Jan 13, 2010 at 2:27 PM, Simon Laws <[email protected]> wrote: > On Wed, Jan 13, 2010 at 2:01 PM, ant elder <[email protected]> wrote: >> On Wed, Jan 13, 2010 at 1:49 PM, Simon Laws <[email protected]> >> wrote: >> >>> The default factory code, as it appears in the spec, seems to assume a >>> single domain per JVM as the domainURI plays no role in selecting the >>> factory instance in the default case. Was that the intention? Seems >>> restrictive compared to what we had originally envisaged, for example, >>> suppose a client wants to talk to services from more than one domain. >>> >> >> That doesn't necessarily mean a single domain per JVM does it, as the >> factory impl itself can still use different domains based on the >> domain URI. It does seem like right now the OASIS code assumes one >> discovered factory impl per JVM, which i thought wasnt the intention >> (thats what I just posted about on the other thread about the client >> api). >> >> ...ant >> > > Well it could guess I suppose but AFAICT the factory impl only gets > domain uri information at construction time. Once constructed you're > into getService() which doesn't take domain uri information. > > Simon >
Ah yes i see. Really the SCAClientFactory defaultFactory static should be the factory class _name_ not an instance so that SCAClientFactory.newInstance really does return a new instance. ...ant
