At 9:50 AM -0700 5/26/09, Lisa Dusseault wrote: >That's an excellent question, but I think like so many others it has >to fall under the judgement of the person writing the implementation >report. Is it OK to just test 2 implementations or is it important to >test 2 servers and 2 clients? It might be possible to go to an >interoperability forum and test 15 different implementations, yet if >that's a protocol for which there's a sizable *other* community that >doesn't implement a required feature, that ought to be noted in the >implementation report. > >I'm hoping that by putting the onus on the writer of the report to >carefully characterize interoperability, that we can encompass many >such judgement questions. On the flip side, if we tried to address >every such judgement question, we couldn't possibly foresee every >corner case. > >Do you have any suggestions for criteria that could be broadly >applicable and useful? One possible criterion would be "if you considered multiple implementations but focused on a subset for the report, at least discuss the wider set somewhere in the report". That could include a simple list of all the implementations considered, but also might include reasons for subsetting (some of the ones left out had difficult admin UIs, had licensing issues, and so on). --Paul Hoffman, Director --VPN Consortium _______________________________________________ Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf