Opened 7 years ago

Closed 2 years ago

#8356 closed task (wontfix)

Figure out how much would change if we compute flag thresholds from bwauths

Reported by: arma Owned by:
Priority: Medium Milestone:
Component: Metrics/Analysis Version:
Severity: Normal Keywords:
Cc: karsten, mikeperry Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

In #8273 we're about to switch over the Flag, Guard, etc computations to look at measured bandwidth, not declared bandwidth.

This is probably good for security in some way we haven't fully fleshed out (or at least I haven't).

Long ago, we had plans to do exactly this, at least for the Fast flag, but we stalled them for reasons I don't remember. Mike?

Karsten, it occurred to me that you'd be in a good position to answer the question "how different would the flag assignments actually be, if we switched?"

Somewhere in here we will also want to answer the question of "what replacement values should we use for parameters like AuthDirFastGuarantee, AuthDirGuardBWGuarantee, whatever Nick called the value of 4KB, etc."

Child Tickets

Change History (5)

comment:1 Changed 7 years ago by karsten

Sounds like a fun analysis. When do you need an answer?

comment:2 Changed 7 years ago by nickm

I want to merge #8273 into 0.2.4, since it's security-relevant. So I don't need precise figures now, but it would be good to know pretty soon whether the effects would be disastrous.

comment:3 Changed 7 years ago by karsten

Okay, looking now...

comment:4 Changed 7 years ago by karsten

Oh well... I'm afraid there's no quick-and-dirty way for me to find out if effects would be disastrous. I tried feeding my old #2911 scripts with recent consensuses, but I'd have to rewrite large parts of it to consider advertised and measured bandwidth at the same time and to include the Fast flag in results. That code runs quite slowly because it's really a lot of data, and that makes debugging harder. Nothing of this is impossible, but I really need to focus on sponsor F today, tomorrow, and in the next two weeks, or we'll run into trouble there. I can't really promise when I can look into this ticket. Sorry.

Of course, you could try what happens if you run the new code on one of the authorities and see if https://metrics.torproject.org/consensus-health.html#overlap shows something unusual. As you can see, there's always some disagreement between authorities. If a disaster happens, you should notice by the authority running the new code assigning very different Stable and Fast flags than the rest.

comment:5 Changed 2 years ago by karsten

Resolution: wontfix
Severity: Normal
Status: newclosed

This ticket has not seen progress in almost 5 years, so I'll close it now due to inactivity.

Note: See TracTickets for help on using tickets.