I wish histories like this existed for every topic. I'd buy a book of these.
A short history of database isolation levels, by @narayanarjun:
https://ristret.com/s/f643zk/history_transaction_histories …pic.twitter.com/Cym6tFUmTB
-
-
-
Thank you! Try this one as well: https://ristret.com/s/gnd4yr/brief_history_log_structured_merge_trees … I'm writing a few more in the next months. Subscribe to the blog!
End of conversation
New conversation -
-
-
Postgres’s default isolation level isn't even repeatable read, much less serializable, because they both require the application to be able to retry transactions that failed.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
/cc
@aphyr I guess?Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
An additional fun fact is that not only were the original ANSI SQL definitions incorrect, they were also ambiguous! This was in part because they were specified in English. A Critique of ANSI SQL Isolation Levels shows that no matter how you interpret them, they are still wrong!
-
The graphs used by Atul Adya are also neat. The idea is you have directed edges from transactions to transactions that must occur later. If the graph has a cycle, it's impossible to serialize the transactions. Otherwise you can order the transactions and the history is valid.
End of conversation
New conversation -
-
-
Following up on the end there, I believe that this falls into the same category of concurrent systems that we don't have proper tooling for testing on our laptops. Simulation can give us cheap local faster jepsen for building distributed systems, but for threads we take big hits.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.