Hi, is there some consensus regarding migration from c++11 to c++17 in Bitcoin core? I mean it seems that c++17 is not as widespread as I thought. So the migration will happen in medium term horizon, right? Like maybe in a year or two, right?
Curious - is there a compelling reason against pulling out all the indexes into a separate codebase/runtime, perhaps also in a repo under the bitcoin org? Only exception I can think of would be that `getrawtransaction` rpc would need to be moved into that codebase for txindex=1, but is there any other instance where existing core behavior relies on the existence of those indexes?
Separating out the indexing stuff would be a good candidate for rust use, and would let us add indexes we've been debating about for a while (e.g. address index) without feeling too bad.
This issue of the message handling thread being potentially blocked by indexing (BIP157/158) that jnewbery raised is disconcerting, and separating out the indexing stuff seems more or less in line with the spirit of the process separation work.
The main downside to have it outside the main codebase is performance IMO
And there are already projects doing the index "outside of the codebase" (like ElectrumX, etc.).
jonasschnelli: yeah, I agree - just might be good to have one that's clearly maintained by the bitcoin github org
jamesob: I think a way would be: start in your own personal account (or a new entity), find (a) capable maintainer(s), find contributors. If there is acceptable development performance, migrate it to the bitcoin-core github organization.
jamesob: we could start by giving them their own process? With #10102 you could then swap out the indexer process for your favorite Rust tool.