2003/12/19 NMWG Request schema phone conference ================================ Last week there were two workshops: 1) A bandwidth estimation workshop 2) Matt went to a performance measurements workshop for the US - they identified a bunch of interface we should all be talking - the first interface is between system & tools 1) AAA 2) ??? 3) request 4) response - incoming message now; summary from Matt later Onto the doc: Splitting schemas ----------------- Dan's assumption presupposes that the test request and the result request can be split at the same place as results in the past and in the future. Eric: This whole issue seems pretty fundamental. Result request is an implicit test request (in the case where a result isn't already available). TODO: Try to give some examples where having a single schema is a bad plan Requesting what characteristics are available --------------------------------------------- Tanya's request. Mark: this is kinda out of our scope Eric: at the workshop last week, this same question kept getting raised. Chaz had the question: how do you ask if something is going on at some time. NOTEME: that's a damn good question. Quite a good thing to add to a GHPN Monitoring service requirements doc. Eric: can this be done in our schema very easily? Requesting multiple characteristics ----------------------------------- do we bother trying to codify requesting multiple characteristics, where they only have 1 thing in common, namely the start time. We need to come up with a couple of very simple examples. If someone wants lots of metrics, they have to do lots of requests. TODO: Eric's solution is that you just say you can write out the whole thing repeatedly. We should ask Tanya. Multiple sources or destinations -------------------------------- Dan's point that we can't do multiple sources and destinations is true. But so? The report schema doesn't support that either. TODO: We want an example from Dan, really. Maybe the same solution as for Tanya's point will work. Time stuff ---------- Happy to split time tolerance into two parts. Min/max: Eric: ditch the start/end time thing, and just do time+tolerance, then do min/max stuff. Seems very reasonable to me. Or do, time+tolerance, then say all results or just the closests. TODO: think about this. I think it's good, Eric's not too attached to it :D Crontab-style times. Argh! Lokik: isn't this just a repeating request again? Same as before. Eric: didn't we have a testingInterval here before? Maybe we can just ask Dan to think how to do that. With testingInterval, we can do anything except first of the month. After the 5th, we should reconvene. Mark or I write up where we've got. Maybe we need two calls that week as we were meant to be getting something back to the NMWG. How about the 6th? Same time.