Ellis Michael

[Ellis Michael @rellismichael]

@Dev14e Would strongly recommend reading Hamlet.

But even then, you won't get everything without at least PhDs in Shakespeare and Early Modern English Lit. Tom Stoppard is a mad man.

[Ellis Michael @rellismichael]

@danrkports We could call it "The Marketplace of Ideas"

[Ellis Michael @rellismichael]

Verifying myself: I am emichael on https://t.co/6aEwmiisxk. Rwl_MNcAogli88uzWb89beKdZbKsuplmoxbW / https://t.co/O5Br2gKE9v

[Ellis Michael @rellismichael]

@lindsey @shachaf Good question. I hadn’t realized it wasn’t reflexive. I think the answer is vector clocks though! Look at lemma 2.4 and the way the define sup on vectors. That would actually be uglier to write with the alternately defined C(e).

[Ellis Michael @rellismichael]

@lindsey @shachaf I agree. I think making -> irreflexive is a mistake. It complicates the checking of vector clocks for ->, for instance. Obviously, it’s never going to be that big of a deal to tack on an extra clause, but still!

[Ellis Michael @rellismichael]

@openargs Penny for your thoughts on Epic v. Apple? Implications of an Epic win seem huge; no idea on the chances. It is the first lawsuit I've seen with its own trailer, though.

https://t.co/heUaX17O0G

https://t.co/XwZel5EHqm

[Ellis Michael @rellismichael]

@dave_andersen Non-locking carabiners, non-redundant anchor, harness doesn't look doubled-back. Horse is gonna die.

[Ellis Michael @rellismichael]

@kellogg_martin When assembling the team for this, did you just pick the next 4 people in your contacts sorted alphabetically by first name?

[Ellis Michael @rellismichael]

@heidiann360 @danrkports @VLDB2020 Right. If a switch and replica can bilaterally respond to a read, and you want the protocol to tolerate their failure, then they can serve stale reads. In some sense, the best you can do is rely on clocks.

[Ellis Michael @rellismichael]

@danrkports @gushfsmith @luisceze New protocol: proof-of-fish. Resilient against double-spends as long honest fishmongers control >50% of the fish.

[Ellis Michael @rellismichael]

@danrkports @gushfsmith @luisceze Nah, @luisceze would put the blockchain in fish (DNA), not the other way around.

[Ellis Michael @rellismichael]

@lindsey @danrkports @wilcoxjay It popped up thanks to @danrkports, but now you’ve given me an idea...

[Ellis Michael @rellismichael]

@lindsey @danrkports @wilcoxjay And yeah! It’s pretty neat. It counts down to deadlines and shows birthdays, too. At one point, it showed how many hours grad students had spent on Steam in the last two weeks. That feature has been deprecated.

[Ellis Michael @rellismichael]

@lindsey @danrkports @wilcoxjay Not necessarily. You can then induct on proposal numbers. The example of the very next proposal number is just the base case.

[Ellis Michael @rellismichael]

@lindsey @danrkports @wilcoxjay When 3 gets prepared, it might see 1 or it might see 2, but either way the value it picks will be the value in question, v.

If there were a gap, (say, we saw 1 and 3 and are thinking about prepare for 4), that prepare might see 1 but also (2, v’).

[Ellis Michael @rellismichael]

@lindsey @danrkports @wilcoxjay If we have a quorum, we know that for all subsequent proposal numbers, at least one server from our quorum was involved in the prepare. Consider the very next proposal number that gets prepared after our consecutive proposal numbers. So, if we saw 1 and 2, consider proposal # 3.

[Ellis Michael @rellismichael]

@danrkports @wilcoxjay @lindsey So, if one of the proposers (say, during the first phase) sees that out of the 3 acceptors, one accepted (v, 1) and another accepted (v, 2), it can immediately conclude that v was chosen. No need to try again with a higher proposal number.

[Ellis Michael @rellismichael]

@danrkports @wilcoxjay @lindsey Just now seeing this (thanks to monitor displaying tweets in the lab).

You actually can optimize this in a real way. Namely, you can say that a value v is chosen if there is a majority accepting that value with *consecutive* (not identical) proposal numbers.