Signal NEWNYM vs Stream isolation

[iguanaaaaaa]

While researching Whirlpool and its implementation of identity separation (sparrowwallet/sparrow#1328), I came across an interesting Stack Exchange question: Signal NEWNYM vs Stream isolation. It was both asked and answered in 2018 by @nopara73. The conclusion it came to was that stream isolation alone doesn't provide complete separation of the identities. However as far as I can tell, this discovery has never been incorporated into Wasabi code. I couldn't find a single issue mentioning this either. Is there a reason why not? Are there plans to include it in a future release?

[nopara73]

Thanks, I've revisited the issue and it was supposed to be fine other than a change that happened a year or two ago.

As you read my answer, I identified 3 things newnym does in Tor's source code:

1. circuit_mark_all_dirty_circs_as_unusable();
2. addressmap_clear_transient();
3. rend_client_purge_state();

1. As you see from my answer is not an issue, it's just what newstream does.
2. As you see from my answer is not an issue IF hidden service is used. However, Wasabi changed to a non-hidden service a while ago due to Tor hidden services getting DDoS attacked and rendered unusable. So this warrants further investigation. Created an issue for that: Change back to onions or investigate addressmap_clear_transient() of Tor code #12734
3. Although I acknowledge it to be a theoretical issue, I base that opinion on this comment where even the commenter says it's far fetched.

In summary, Wasabi should change back to onions if the DDoS is fixed or investigate what addressmap_clear_transient() is doing. I didn't bother back in the day because I was going to use hidden service to begin with.

[iguanaaaaaa]

Fair. Thanks for the detailed response!