Huh – I just noticed that Tumblr’s post IDs jumped six hundred quadrillion places on February 24th of this year. It happened at some point between 6:20 PM and 6:35 PM UTC, from the look of it. I wonder what that was about?
Tumblr switched from a sequential ID system based on ticket servers to a system based on snowflake IDs. To preserve sortability, Snowflake ids use the uppermost bits of a number to represent the timestamp of the post’s creation and the lowermost bits to represent worker ids and random entropy. This means they’re going to starting near the limit of 64-bit numbers, several orders of magnitude above where tumblr’s id space was beforehand.
The well-known implementation complexities of using snowflake-based systems with javascript’s 53-bit numbers was the cause of liking and reblogging being broken on the desktop Tumblr dashboard for the majority of the 24th: https://tumblr.zendesk.com/hc/en-us/articles/360043298974-February-24th-2020-Intermittent-errors-affecting-Tumblr
Tags:
#The Great Tumblr Apocalypse #Tumblr: a User’s Guide #the more you know
Brinens and Things - Tumblr Mirror 