soo I was studying about asmap-data, and studying the data from the repo, it comes out that the top ASNs – AWS, Lumen, Cloudflare, etc. appear over 30,000 times. there’s probably a mismathc between the encoding strategy and the data distribution since the _VarLenEncoder forces 16 to 24 bits per reference. Given that we have the top 256 ASNs accounting some 88k occurences (21.2%), isn’t the 16 bits just the waste of extra 8 bits for the storage ? possibly in the coming years, given that BGP tables are growing at a rate of 4-8%, they will outgrow the ~2mb mark that Peter Wuille gave in PR #16702 of the bitcoin core. will this cause problems if pushed to embedded devices ? like any cache evictions or OOM crashes. just curious to know what the designers and developers think. this is probably a very small issue/improvement to work on anyhow i guess…
