Dňa 14. augusta 2024 13:48:38 UTC používateľ Dave Crocker via mailop <mailop@mailop.org> napísal:
>Making a distance-sensitive assumption about traffic behavior is a suprisingly >bad idea for anything having to do with the Internet. Resources and their >uses can be -- and often are -- a long way away and using connections with >wildly varied performance characteristics. Yes, you can be right with distance, if we would talking abou timeouts in tens/hundreds of ms. But we talk about 300 sec vs. 30 sec timeout for **TLS handshake**. Eg. Nginx (web server) has default TLS handshake timeout 60 s. In TCP Hanshake, by default on nowadays Linux, last SYN+ACK is repeated after 63 s... Thus both are far from SMTP's 300 s... We live in 21. century, we (people) are inventing TFO, because 3-way TCP handshake is too slow. We are inventing TLS Resumption, because full TLS Handshake is too slow, or even TLS early data. We are inventing HTTP over UDP, because TCP is too slow. We have Gb/s connections, we are saving RTT, calculating every ms... Anyway, i understand original suggestion not as distance in meters, but as distance to contact... And that distance is definitelly different for SMTP and for Submission(S). >The only time distance-sensitive assumptions might make sense is when the >interaction is known to be -- and must be --- part of a LAN that both parties >are on. Maybe. No, only reliable distance assumptions are on loopback. ;-) regards -- Slavko https://www.slavino.sk/ _______________________________________________ mailop mailing list mailop@mailop.org https://list.mailop.org/listinfo/mailop
