I found this a bit interesting:
https://www.usenix.org/conference/nsdi18/presentation/gengIt describes an algorithm for "nano-second" precise timing using network peers, almost seems to good to be true.
Timing is important for blockchain technology.
PascalCoin git log does contain some hint/reference to a possible timing attack and a safeguard against it, where double spends or other attacks might be possible if time is off by too much. Saw a website about it describing an attack where a single system gets dossed or something like that and then fooled into believing a fake chain and such.
Anyway PascalCoin since it's P2P seems like a nice candidate to include this advanced algorithm. Currently I have no idea how it works =D
...
=D
(Gonna watch this video seems interesting and kinda bored right now =D)
Ok watched the video, the idea is very simple and pretty cool.
I wonder what kind of accuracy/precision PascalCoin and other blockchains need. The "attack" document did describe that even a clock sync off by 1 second could be exploited, so the answer is probably as low as possible. Kinda funny idea this video, highly recommend watching it.
(Didn't know nics could do timestamping, may look into it. Idea in short is to send two packets, and measure time between them, any difference as recorded, signifies queue/noise delays, sometimes they will have zero delays and no noise, so idea is based on "filtering out" these perfect two "packet pairs" which are called "probes" or "coded packets". I do see a flaw in this idea though, if these "ideal" pairs never occur then this system won't work
anyway once these ideal packets are filtered out they are put through something called SVM, some kind of machine learning software, never heard of this before. And a lower/upper bound can already be determined I think even in noisy channels, the filtered probes, cut back noise even more, and finally SVM can find the "red line" between them or so... I will google this SVM to see what it is about. Anyway the idea is interesting, but there could be other technique to filter out this noise
It later goes on about some kind of "circle loop" mechanics to "label" / "time" loops ammong many K servers, and finally all results collected on one chosen server/"master" and
finally re-distributed. Would be nice if this phase/step of the algorithm was more "distributed" maybe that possible too not sure).
Kinda wonder if PascalCoin has it's own "Network Time" detection ? I did see some "Network Adjusted Time" but perhaps this is something different ?
I think bitcoin does try and determine a "consensus on what the time is by some p2p algorithm or so". This may be attackeable but so far is holding up.
Also this new "huygens" algorithm is new, and has probably not been examined how "temper" proof it is. What if clients/servers start to missbehave ?! Hmmmm.... might be useless algorithm not sure about that
Here probably most "voices"/"votes" matter and might make it temper proof against maybe "idealizing" probes or whatever or "fraudelent reports to master".
(If implemented in PascalCoin, this could be PascalCoin's first "Distributed APP" an APP to determine time to 10 to 100 nanoseconds precision or so.)
