> If we’re discussing host based versus network based happy eyeballs, would > it be naive to think that the network based HE would interfere with the > client’s HE? >
Currently this draft only considers IPv4/IPv6 racing situation. The general address racing is already done for all clients (rfc6724). As mentioned in the draft, NHE can work independenly(without interferance) with the adoption of Client-side HE. If client's already support HE, it will query A & AAAA at the same time. NHE can help reduce the unnecessary traffic emitted by HE client becuase the AAAA record will be omitted or delayed if IPv6 connectivity is poorer. I don't see any interferance now. > A router knows very little about end to end properties of a connection. It > could of course do those measurements by looking deeply into packets, but > it would still be restricted to it’s own topological location. > It depends on how far you do the measurement to simulate the end users. Some APM (application performance measurement) vendor like New Relic provides sdk implementing on Apps to provide measurement data for analysis and deliever a good performance monitoring and traffic engineering. Some provide probes close to clients and edge. I think maybe there is a crose-layer issues on this draft which should be simple enough and aim IP layer issues. But right now lots of complected application layer complexities (like CDN, APM) are involved. > Compare that with the data available to an MP-TCP host stack. > I think MP-TCP or other transmition protocol are based on stable, high-performance ip layer. > And note I think HE can’t just be between v4 and v6, but between all the > candidate connections between source and destination. > HE(RFC8035) only focus on IP connection seletion right now accroding to the problem statement: "Many communication protocols operating over the modern Internet use hostnames. These often resolve to multiple IP addresses, each of which may have different performance and connectivity characteristics. Since specific addresses or address families (IPv4 or IPv6) may be blocked, broken, or sub-optimal on a network, clients that attempt multiple connections in parallel have a chance of establishing a connection more quickly. This document specifies requirements for algorithms that reduce this user-visible delay and provides an example algorithm." If there are other candidate connections, they also need a better IP connectiviy as a bais I think. NHE at least can give a indicaiton of poor connection for certain address family (or address) . It is potentially a low level routing founctions other than hight level traffic scheduling tools. Davey
_______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop