Re: [IPFIX] review of draft-scholz-ipfix-rtp-audio-quality-00
Andrew Feren <[email protected]>
| Newsgroups | gmane.ietf.ipfix |
|---|---|
| Message-ID | <[email protected]> |
If in doubt just go with the unsigned64. It is always possible for an implementation to use reduced length encoding and send fewer bytes if that implementation doesn't need all 8 bytes. -Andrew On 08/06/2012 08:14 AM, Michael Krüger wrote: > Hi Hendrik, > >>>> 3.2. rtpMOSClass1 >>>> >>>> Description: Number of seconds the monitored stream had a MOS >>>> quality lower than 3.10 >>>> >>>> Data Type: float32 >>> It's unusual to represent time as a float in IPFIX. Could you pick a >>> suitable resolution (eg, milliseconds or microseconds) and export an >>> integer? >>> >> The purpose of these IE items is to report the amount of full seconds >> VoIP streams have been in either one of these 5 MOS classes. So I >> suggest to change the type to unsigned64. >> >> Regards, >> Michael >> > I did some math and really believe this should be an unsigned 64-bit > value. Since the purpose of this IE is to provide an aggregation over > all flows carried e.g. over a SIP trunk / carrier-interconnection the > worst case would be potentially many millions of flows aggregated. > Reporting the quality like this for a single flow is possible, but > likely not the general application. > A 32-bit resolution to report quality distribution of full day would > only be sufficient for 49k flows or 24k bi-flows / calls. That doesn't > seem to be a lot for quality reporting on a daily basis on a > carrier-interconnection. Therefore unsigned64 with units = seconds seems > for me the way to go. > > Regards, > Michael > > _______________________________________________ > IPFIX mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/ipfix _______________________________________________ IPFIX mailing list [email protected] https://www.ietf.org/mailman/listinfo/ipfix