That's the point James that's your opinion and even with that worst case scenario of volume of requests - whilst not a great use of CPU and I/O, it's not that great a technical challenge.
On 2 Dec 2016, at 10:43, James Harrison <ja...@talkunafraid.co.uk<mailto:ja...@talkunafraid.co.uk>> wrote: On 02/12/16 09:49, Neil J. McRae wrote: In my view that's not what is expected. Of course, it's all an opinion until tested. I suggest common sense applies until someone says it doesn't. It may not be what's expected, but it's what's possible. Or are you going to try and guess what's all really part of the same stream of communication? How do you actually do that in an auditable fashion (because HMG will want proof this is being done consistently and correctly, right?) and in a way which scales to the volumes of traffic that ISPs deal with? Common sense from a technical perspective says that each individual connection requires an internet connection record, and that makes the technical challenge at least reasonable in definition. If we assume a single connection per MPEG-DASH segment and 15 second segments, your average movie just generated 480 ICRs over 2 hours. I think that's probably not the worst problem - the issue will be the average web page/Facebook session creating dozens to hundreds of ICRs per page load, every page load. Popular point to point IM/WebRTC systems like Discord etc might also be unwelcome culprits adding a lot of short-lived UDP traffic bursts.