Testbed Support for GridPP member institutes [mailto:[log in to unmask]] On Behalf Of Daniela Bauer said:
> I think we should roll back glue2 which is way overengineered and stick with glue1.
You can argue about whether it was right to develop GLUE 2 at all, but in practice GLUE 1 has had no development since about 2008 and isn't likely to get any, so if the information system is expected to have any kind of future GLUE 2 is really the only option. You could just keep everything as it is with a view to phasing it all out soon, but even then switching off GLUE 2 would lose you information, e.g. the OS version under a service (as distinct from the one on the WNs) and the Nagios monitoring only exist for GLUE 2. Also ARC and the GOC DB have adopted GLUE 2 as an information model. There's also still work on it in the US for XSEDE even if OSG aren't interested (I think their JSON rendering spec is in public comment right now, and XML was approved fairly recently). Basically the GLUE 2 language, i.e. the object and attribute naming and definitions, is the only common way to represent these kinds of concepts across different systems, regardless of whether the BDII itself survives.
Stephen
|