Here is a conversation with nopara73, waxwing, aruk and I.
<nopara73> @belcher There is something I don't understand. Why not just run an Electrum server on your own machine and point the Electrum client to localhost?
<belcher> ill write an FAQ with that since that question has come up a few times
<nopara73> I get it that pruning and stuff doesn't work like that.
<belcher> the answer is because a normal electrum server needs lots of disk space and memory
<belcher> yeah thats it
<belcher> an regular electrum server has a database of every bitcoin address that was ever used
<belcher> this gateway thing only has your own addresses, and they are stored in bitcoin core's wallet.dat
<nopara73> However instead of building a completely new system, for me it seems like the first reasonable approach would be try to tweak the existing system into being more space effective.
<belcher> theres tweaking and then theres trying different approaches
<belcher> storing every address ever used is inherently not scalable, storing only your own addresses is better for some things
<belcher> the tradeoff is you have to -rescan if you import addresses that have historical transactions
<nopara73> I see, makes sense.
<belcher> thats why this probably wont be added to electrum itself because it has "instant on" as one important value
<belcher> this isnt "instant on", it trades away that in return for full node verification and less resource usage
<nopara73> Oh, there's something else in my mind.
<nopara73> What do you think of the future of on-chain transactions?
<nopara73> It seems like they are going to the roof.
<nopara73> And the situation will not get better.
<belcher> they'll be limited
<belcher> my view is basically the same as bitcoin cores/UASFs
<belcher> block space is a scarce resource and we'll have to learn to use it efficently
<nopara73> This may comes with a trend of people who are using Bitcoin on-chain will also be easily able to afford to run a fullnode.
<belcher> for privacy specifically i wrote this
https://www.reddit.com/r/joinmarket/comments/7m422w/some_updates_on_joinmarketclientserver/drtzm84/<belcher> privacy will have to happen off-chain, because buying lots of block space with coinjoin/whatever will be expensive
<belcher> reading it again now, i should put chaumian ecash servers as another off-chain tech and link your blog post
<nopara73> Recently I was wondering if building a nice UI on top of fully validating Core is a strategically smarter decision than trying to tweak with light clients.
<belcher> yes i think so, i really like electrum's UI so my gateway idea is about having both electrum's nice UI and core's full verification
<belcher> in your bitcointalk reply you say "Building a nice UI would be less of an effort then building this gateway." which is... not true... building a nice UI means re-doing all the work that the electrum devs have done
<waxwing> yes agreed, it's not really even quite "nice UI" that electrum client has got so right, it's feature-richness imo.
<belcher> anyway if its okay with you id like to copypaste this conversation into that bitcointalk thread
<waxwing> feel free to repeat anything i said.
<nopara73> @belcher My point was building a UI is just putting one thing on top of another, while building a gateway requires actual engineering and logic. Maybe building a UI would be more work, but it would be an easier work, where you don't have to be anxious about if your algo fails, people can lose money.
<waxwing> @nopara73 that's true, but he doesn't want to change the UI, he wants to change how easily you can replicate the electrum server functionality backend as a single user with one node
<waxwing> you can do it now but it's too expensive (txindex and so on)
<nopara73> Ok, but that's just a personal engineering boner, not an actual, "what's the most effective way to fix that problem I have: decision:)
<waxwing> you think so? i don't. if you scroll up you'll see i made the same point as you about reusing existing arch; but he intends to create a far lighter-weight version of it, which would address a specific need: to use electrum UI without a heavyweight server backend, for a single user.
<nopara73> Indeed. I didn't understand what you were talking about, because when I read your comments I didn't yet read the bitcointalk post.
<aruk> It would actually be very useful as using electrum server with your full node is basically the only way to validate your hardware wallet transactions with a full node
<aruk> I wonder why this is not a more wanted feature
<belcher> aruk yes that is one big motivation, i realized we talk up hardware wallets as a great way to store privkeys, but bitcoin security also requires using a full node wallet and hardware wallets dont work well with those (yet)
<belcher> for engineering vs UI, i dont like UI that much (personally)
<aruk> it does but it's extremely cumbersome
<belcher> i actually liked reading cs textbooks with all those graphs and heaps and tree diagrams
<belcher> creating UI i mean
using it is obviously nice
<waxwing> only thing is, like with JM, if you go with a non-indexed approach, you have problems in recovery.
<belcher> yep, tradeoffs :\
<belcher> its required for scalability i think, transactions come in blocks, you have to scan them
<nopara73> @waxwing My point is you get the same result if you replicate the Electrum GUI for, Core, or let's say @jonasschnelli’s fullspv PR here:
https://github.com/bitcoin/bitcoin/pull/9483 then it's an easier way to get to where he wants to get, than writing the gateway.
<belcher> you could make a database of every address every used... but that will be big and probably centralized because of that (i.e. people will use bc.i's API instead of scanning their own blocks)
<belcher> nopara73 jonasschnelli’s PR doesnt support (for example) mnemonic phrases
<belcher> or any other of electrum's nice features
<waxwing> yeah it isn't just GUI, like i said above, it's feature-richness.
<belcher> yes, UI is the wrong word
<waxwing> HW wallet integration mentioned above is a very good point too.
<waxwing> several other things, but i don't want to be unfair to Core's GUI since i never used it :laughing:
<belcher> features are a part of UI i guess... /me looks up how "UI" is actually defined
<nopara73> What? Why doesn't it? I support it in HiddenWallet. It's just the user must specify a date when the wallet was created.
<nopara73> At recovery
<aruk> if you want consensus to be enforced, you need people validating high value transactions. Nobody is doing it because they all use a hw wallet that does not depend on a local node
<belcher> Core's GUI doesnt support lots of stuff, i think most core developers are interested in the super-important scalability stuff
<belcher> there's a huge amount of work gone into scalability in bitcoin core, without which bitcoin would likely die
<belcher> IMO the best application in Core is bitcoind not bitcoin-qt
<belcher> the way i run things is an old always-on laptop with bitcoind running (and other stuff like joinmarket and irc)
