Jeff Tantsura at our IETF117 presentation mentioned that it was not clear how SCION does in terms of convergence time.
Then we also got this question from the iSE review:
From: Joel Halpern
There appears to be a very fundamental gap in the scaling description. On the one hand, SCION claims to scale well (presumably to large numbers of participating AS. There is no discussion of the effect of the number of participants or meshiness on the number of PCBs. There are very strict limits on the rate at which PCBs are propagated. But no analysis as to whether the system will ever converge . Particularly if the number of interfaces between AS becomes large.
Something about convergence time and scalability should be added to the draft. We could start from something like this:
Note that, since there is no convergence phase in SCION, we cannot compare to BGP’s convergence time. SCION path-segments are stable as soon as they are disseminated.
From Roger Lapuh: Explain why why 4030 ISDs is enough for public addressing. Refers to section 2.2.3 "Addressing > ISD Numbers"
From Juan Garcia Pardo: Where do the suggested values for set size and propagation time of PCBs in section 2.3.1.1. "Storing and Selecting Candidate PCBs" come from? Do they depend on the network topology? Explain.
Jeff Tantsura at our IETF117 presentation mentioned that it was not clear how SCION does in terms of convergence time. Then we also got this question from the iSE review:
Something about convergence time and scalability should be added to the draft. We could start from something like this:
Source: https://netsec.ethz.ch/publications/papers/2021_conext_deployment.pdf
(Copied from https://github.com/scionassociation/standards/issues/82)