Pls explain

It's kind of dumb that ~everybody is running an electrum server to use it as a backend for their wallets. Uptake on blockfilters is pretty small in comparison.

Ah, right.

I don't think most users need a full address index, since they usually only have a handful of wallets with a dozen addresses used. It's so wasteful having an extra 100GB for an index you'll lookup 0.001% of it's entries I've proposed to the Electrum guys replacing script hashes for script pubkeys, that way, using an EPS-like setup gets way easier.

Yeah it sucks, but if most people that run some form of home node setup do that already anyway, I feel like we should confront that reality. People want to import existing wallets and not wait hours for rescans to finish.

With CBS and height hints it can be minutes, tho. My goal with the new endpoints was to make it possible to use CBS for electrum servers. But sure, that would need more work on the community side.

Would you add an index to libbitcoinkernel? I thought I heard that somewhere

No, that is not what the kernel lib is for. One potential avenue me and some others are exploring is using the library as a shortcut into bitcoin's internal block store to make indexing faster. We are working on isolating the indexes into separate processes, which makes them more programmable and fault tolerant, so a bug in the indexes doesn't crash the entire node.

Ah, that makes more sense. Separate those concerns.

I implemented this once in bitcoind, once with nodejs and another time in bcoin. The architecture in bcoin with plugins I think was the best and could run in a seperate asynchronous task (not part of consensus code), it could also be extended to index (or count) anything, types of scripts used, fees and other info.

We also used plugins for blockfilters, if I remember correctly.