vtnerd moneromooo: I was really hoping to replace the macro hell of epee with json_dto, since it mostly implements the same thing, its API is nicer, and appears to be a lot faster than whatever epee is doing. I could check to see if your PR makes a difference?

We can initially rewrite the KV_* macros to generate the json_dto serialization, to make the transition easier

is it possible today to run a monero node fully over tor? not just tx broadcasting

Tor bandwidth seems pretty decent?

I thought the issue was latency, not bandwidth. fluffyblocks seems to address the bandwidth issue already

the idea being that no monerod-related traffic is seen from the IP

i haven't visited this issue in awhile, with bitcoind it's possible to have everything over tor, it finds .onion peers etc. I'm wondering if it's the same with monerod these days, iirc only broadcasting of submitted txs was implemented, not block sync .. but I could be wrong

hyc: I agree. But also absolute bandwidth seems to be at decent levels. Latency, not so much.

it's fine, the delays are usually in the 300ms range

300-500

for a monerod/bitcoind node I don't see why that is a problem, for a gaming server or furiously fast stock trading, sure

there is a paper where bitcoin nodes can be IDs even over tor because of the distinct timing

Tor is no panacea

hasn't the tor project taken steps to avoid the timing issue, they'll have you send dummy data all the time to avoid that?

"macro hell" is definitely not a good argument to me. It works. Rewriting jsut because you don't like macros is a non starter because next guy will want to rewrite because he doens't like whatever you use instead of macros.

I like faster building and faster execution. I don't really care about binary size I think (until shown how it matters for embedded targets). I care about changes being small.

So if vtnerd's comment about getting most of the speedups with small changes works, this seems much better to me.

Fair enough. I'm not vehemently opposed to macros, that's why I suggested implementing the the change-over by changing the current macros to generate json-dto code

But all those efforts would be in vain if vtnerds changes are better, so I'm curious to see how that performs

there is also something to say about using things which are out there instead of reinventing the wheel all the time, gives more confidence that it is maintained and works those are also attributes which you want to have and should not be ignored