renghardt
commented
5 years ago I just went over my notes from IETF 103 again, and I noticed three more points that perhaps make sense to include in the path selection use case:
- Such path selection might hope to be "better than random".
- Path properties may come with limiltations, e.g., performance-related path properties measured in the past may become outdated and not related to path properties of packets sent now.
- There are trade-offs between different performance-related path properties, e.g., latency and data rate. What constitutes "good performance" also depends on the application's needs.
This is quite a lot, I realize. I can also put in these parts later in a new PR. I'm just trying to incorporate as many IETF 103 comments as possible before going back and asking the Research Group whether we are on the right track.
Motivation for why we need the path properties we decided to list. What do you think about this list?
So far I only added three systems that can use the path properties, but did not specifically say which properties are relevant for which system (e.g., achievable data rate is relevant for p2p file sharing systems). I'm not sure how much into detail we should go here...