Victor Grishchenko
Victor Grishchenko
...the size of the resulting frame in RONt is about 100KB. That space is mostly taken by ids. An empty replica named "test" produces ~25byte/op updates in "[B1.4] Insert N...
Actually, the absolute worst case in RONt is 60bytes per a single-character insertion op. Like, if we max out every field: numbers, replica ids, non-BMP Unicode chars.
> Do you think it makes sense to add Swarm to this benchmark, as it is based on RON? I don't think so. gritzko/swarm is no longer supported
Hi Kevin! The link is 404 Sent with [ProtonMail](https://protonmail.com) Secure Email. ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Thursday, July 16, 2020 4:52 PM, Kevin Jahns wrote: > Hey [@gritzko](https://github.com/gritzko), > >...
Weight Self Weight Symbol Name 16.23 s 100.0% 0 s RealmTasks (15153) 6.70 s 41.2% 0 s _dispatch_worker_thread3 0x66303 6.70 s 41.2% 0 s start_wqthread 6.70 s 41.2% 0 s...
either the core is called repeatedly or it goes into a loop committing a transaction
maybe @morten-krogh has some idea
a variation of #370 most likely (and #370 is most likely an infinite loop in the app committing something to the db)
An append-only or a sorted container needs no tombstones, indeed. Arbitrary insertions need reliable location ids. It would be helpful to apply the 5 whys technique here https://en.m.wikipedia.org/wiki/5_Whys What problem...
I am not fluent in the dot jargon, but this is how I understand it: you need a scheme where data absence is meaningful. Like, a store has its version...