Holy shit about the collation thing. That’s terrifying.
Not to mention the unexpected differences between similar collations (on red hat awhile back, enus and enau alphabetised Thai characters completely differently). We were inconsistent by accident, and fixed it, but I’d have expected them to have the same idea of alphabetical order.
I hope someday PostgreSQL extends to support sum-type tables as well as product-type tables. If I want to store a record that might be one of seven distinct things, that gets pretty messy—especially if I want a single ID that’s common across all of them.
Table inheritance may be useful in that situation–put the ID in the parent table and the distinct columns in the sub-tables. (https://www.postgresql.org/docs/9.6/static/ddl-inherit.html)
It may slip under the radar for many, but I have so much <3 for this commit mentioned in the article.
I have a pretty strong preference for handling failover at Layer 7 rather than Layer 2/3 (i.e. with virtual IPs). This change makes that way easier!
With all the nosql craze and all, PG is still my favorite DB. Great software, extremely reliable.