On 6/8/2021 9:39 AM, ais523 via agora-discussion wrote: > On Tue, 2021-06-08 at 09:22 -0700, Kerim Aydin via agora-discussion wrote: >> I had a feeling you'd bring this one up when I saw you register and saw >> that your computer delivery was faster than mine :) ! > My CFJ earlier was actually primarily intended as a test of my delivery > times to -bus. I decided to disguise it as something other than an > email timing test, because I think that if I'd made it obvious that I > were planning a timing scam, someone would have figured it out and > might have pulled off the counterscam.
I've definitely done that in previous timing races. FWIW, this time I made the conscious decision to push it to where it would likely arrive late. This was not to gain the advantage of "no reaction time", but on the off chance that Aris et al. decided to distribute a final proposal that I wouldn't otherwise be able to respond to, I wanted my extra strength to have the greatest chance of being spent after that proposal's voting period started. > I think the best choice (which might require a rule change) would > probably be along the line of "an email is sent at the point in time at > which the person sending the email completed the process of telling > computers to send it, except if they introduced an artificial delay > into the process". Instead of forcing timing-scamsters into having to > find a tradeoff between ensuring the email arrives on time and ensuring > people can't react to it if it arrives too early, it makes sense to > just say "as long as you aren't trying to rig things, just make sure > you send it before the deadline and don't worry about when it arrives". I agree that this is the only "equal and fair" option I can think of. Though we might have to introduce tiebreakers - if we are permitted to go as close to the wire as we like, CRON jobs can make it exact, so we might expect a few more ties to happen. I can't remember any cases that ever had ties - would messages that were tied down to the second be placed in the right order in the archives based on milliseconds? Or based on one of the receipt timestamps (which would be something of a compromise - if you're expecting a tie you go back to timing your messages and if you're slow gamble on being 1 second early). -G.
