While looking at the JBossWS webserviceref regressions (from AS5 Beta4), I found that the way the webserviceref name is set when no name attribute is used changed.
For instance in this case:

package org.jboss.test.ws.jaxws.samples.webserviceref;
...
public class EndpointClientOne
{
@WebServiceRef
static EndpointService service4;

@WebServiceRef
static void setService6(EndpointService service)

{ EndpointClientOne.service6 = service; }

private static EndpointService service6;

@WebServiceRef(value = EndpointService.class)
static Endpoint port3;

@WebServiceRef
static Endpoint port4;
...
}

we had the following service ref name in AS5 Beta4:
org.jboss.test.ws.jaxws.samples.webserviceref.EndpointClientOne/service4
org.jboss.test.ws.jaxws.samples.webserviceref.EndpointClientOne/service6
org.jboss.test.ws.jaxws.samples.webserviceref.EndpointClientOne/port3
org.jboss.test.ws.jaxws.samples.webserviceref.EndpointClientOne/port4

and we have the following ones in AS5 trunk:
service4
service6
port3
port4

In short the declaring class name was prepended to the name when the @WebServiceRef's name attribute is not used.
Was the naming convention changed on purpose or we can use the same as before? This cause regression because we have the deployment descriptor with the as5beta4 style names and (I guess) when the metadata coming from that are merged with the metadata coming from the annotations... we end up with more refs than what we should have.