Two of the most horrendous issues were systemic address reuse where change addresses were later used again for mix outputs; and symmetric address reuse where the same address would show up on the input and output sides of a CoinJoin transaction. Neither of these issues were resolved with Wasabi 2.0 (Sabi Wabi or whatever they called it) and you can still find these issues occurring in Wasabi CoinJoins today.
If you want to go down this rabbit hole, here is a good place to start: https://pastebin.com/NHDVC03U
As for the new Ashigaru Whirlpool coordinator being able to link inputs and outputs, despite the unhinged takes flying around from NothingMuch and FloppyDic, the claims don't make sense IMO given the description of how the protocol works. If either of those guys have a valid point, it is getting lost due to their poor communication skills and inability to demonstrate or explain the exact issue in laymen's terms. I have seen zero evidence that a ZeroLink coordinator like the one Asgigaru is running has ever linked inputs to outputs.
quotingTruths about coinjoins:
nevent1q…u3p0
- Wabisabi has a superior architecture to Whirpool.
- Including change in the coinjoin as in Wasabi 1.0 is in any case an advantage, not a disadvantage. If you don't understand this, it's simply because you don't understand the math.
- There is no such thing as isolated change in Whirpool; changes are traceable.
- With Wabisabi, you can forget about change problems and Postmix tools.
- Wabisabi is not free as they sell it to you; the coordinator continues to make money from decomposition.
- Wabisabi does manage Tor circuits well; I checked the code myself (Nothingmuch says no, and I have to trust him because he knows much more than I do).
- Whirpool does not manage Tor circuits well, and the coordinator can find out who is on each side of the mix (I have not reviewed the Ashigaru code that they say has mitigated this, proving me right, which I was denied at the time).
- Whirpool and Wabisabi are vulnerable to tagging attacks.
- In Whirpool and Wasabi, you have to trust the coordinator; you don't know if they are really executing the published code.
- Joinmarket is complex for someone who doesn't know what they are doing, but it can mitigate the problems described above.
- Joinstr, in my opinion, is the most promising; it mitigates most of the problems described in centralized coordinators and is easy to use.
I think the community should throw its weight behind supporting #Joinstr