glue-wg-bounces@ogf.org [mailto:glue-wg-bounces@ogf.org] On Behalf Of david.meredith@stfc.ac.uk said:
2) WRT the xrootd I understand that "xroot" was decided upon in Glue2 rather than "xrootd" since xroot is the protocol name and xrootd is the server.
To be a bit more precise, "xroot" was decided in GLUE 1 for the relevant GlueServiceType. For GLUE 2 that generally gets copied to the EndpointInterfaceName, and also to the ServiceType at least for simple services, although I'm not sure if we agreed general principles on what the ServiceType should mean.
However, the requestors of the xrootd service types strongly prefer 'xrootd' since they need to determine the exact implementation which matters to them. Now, I realise that the 'EndpointImplementationName' is meant for that purpose, but GOCDB doesn't support glue2 (yet) and to be pragmatic, I expect many dependent systems within EGI will still require a single SE type list for a long time to come yet. Therefore, what are your recommendations? Would it be possible to un-deprecate the 'xrootd' SE type in the GLUE2 list or should we consider this only as a GOCDB SE > type?
I think a more basic question is how you intend to map your GOC DB type to GLUE 2 in general - Maarten seems to want it mapped to the ImplementationName, but I don't think that's what we've been assuming up to now. We could of course decide that the InterfaceName should be xrootd, it's just an arbitrary name, but still what it would be naming is the protocol and not the implementation, and the expectation would be that anything which presents the same interface would use the same name regardless of what lies behind it. Stephen -- Scanned by iCritical.