So, I today listened to @openoms on the TFTC Podcast : https://anchor.fm/tales-from-the-crypt/episodes/155-Openoms-eddai0
It was great to hear in-depth discussion of Joinmarket :) I wanted to add several points, some just personal opinions, others corrections:
On probably the biggest question the hosts were asking: "what is the economic incentive for makers, why are fees low, is it easy to Sybil", here are several thoughts of mine:
* The market-arrived price has settled in the 0-1000 sat /maker for lower value coinjoins .. (1/n)
.. in my opinion, for a combination of these reasons: (1) the network cost of creating the txs is significant (2) the makers get *a lot*, but not all of the benefits that takers get, as long as they are prepared to wait for a long time. (3) the cost has to factor in hot wallet risk, but this naturally trended down over time as people saw no evidence of JM coins getting stolen. Note that contrary to what was said on the pod, fees *were* in fact much, much higher in the first year. (2/n)
@openoms Doubtless there are other reasons than those 3 but they matter a lot, I think. For larger value coinjoins fees are probably mostly still ultra low, but liquidity limits might result in you paying a bit more in total.
Overall my point there is to suggest that basically low fees are pretty rational because the taker bears the main network fee cost.
On Sybiling, I think you did a good job of covering a lot of the issues, but I hope people understand the change that happened ... (3/n)
.. a couple of years ago that I consider very important. In late 2018 we did something that had been obviously needed since the Moser paper in 2016 made the claim that Joinmarket could be Sybilled for $16K (something like that). I at the time explained why this was kind of half-true at best, but it was anyway observing something important and that was finally addressed in 0.4.0:
random-under-maximum basically sacrifices exact taker price choice for ... (4/n)
... much better Sybil resistance. Now, there are other roads we can go down and one of the most advanced is the fidelity bond idea which is currently being worked on by @belcher . But I really hope people who use Joinmarket understand this mechanic; we *don't* have strong Sybil defence, but we have a more practically realistic defence than might be imagined just from thinking of it as a "fee market". (as well as the point you explained about "all counterparties", of course).
(6/n) Other points: on binaries, there has been no AppImage from me, it seems the message was confused on IRC, although i was pretty explicit (posted exact command lines there): I have simply replicated what I did already in 2016, that is, built a binary with pyinstaller; it just happens to work a bit better now. W.r.t. having dmgs for mac users, I wish we did. I would continue to ask for mac users to help us out.
On the more general point of "it's perceived as hard ... (6/n)
... to use/install", well, it's nice if people at least mention that you can install it easily in one line with Linux desktops, even if that doesn't extend to other environments. I *thought* the install was fairly painless on Mac, at least that's the impression I got.
But all perceptions like this are usually based on some reality :) You probably heard me talk about this difficult question of "easy to use" on Livera. Our app is pretty damn complicated.
I think you conveyed a lot of other practical considerations very well, e.g. switching maker/taker roles, and rules about how the 5 accounts/mixdepths are used.
You also explained the crucial point that Joinmarket takers are basically the only people that can do coinjoins of a payment amount, or coinjoins without change of a set amount.
Of course I wont' comment on all the other content but it was definitely an interesting listen.
So in the end, I would hope the hosts would give Joinmarket a try, especially the GUI, afaik it isn't even hard to run that on a Mac (though hmm, maybe they are Linux users, not sure). They can even payjoin with it today, although the btcpayserver version should be coming soon :)
But yeah this was a really interesting deep dive to hear discussed.
@waxwing @belcher Unfortunately itvia.consider to be hard to install something if there is any command line usage involved.
Most expect the installation be a double click and an OK button.
This I think is hardly an appropriate expectation for JM, since the usage requires grasping some more complicated concepts.
I have tried that .Appimage from IRC myself, but I understand it was only an experiment.
@waxwing @belcher I do think a possible way forward is to include JM in these preconfigured packages we provide with the RaspiBlitz with the modified install process helping with the connection to bitcoin RPC, later password changes and activating the virtual environment too.
Will work on helping the users with the syntax of the different commands and running yg as a service too.
Planning to work on this this weekend on the Fulmo #LightningHackSprint.
Bitcoin Maston Instance