Thanks for the tip. Persuasion isn't my forte and I don't think I've ever seen a pro-schema argument win against a let's-make-table-soup team lead.
It is so frustrating when people inevitably converge on an obvious schema that they just don't want to let the database know about - or develop the Things table that re-implements database schemas poorly. The old table that is clearly several other tables mashed together.
This data is going to be the longest lived and probably most valuable artefact of the company. It is nearly free to put a schema on it, it improves correctness and performance more or less for free. Yet still it isn't possible to get consensus that telling the database the data schema might help.
People don't realise how inefficient JSON blobs are either. I swear there must be AWS databases losing a significant fraction of their IO budget to re-parsing what should be column names with every data point read. I can only hope that postgres devs have rolled their eyes and optimised the case where people don't know how to express tables in SQL.
It is so frustrating when people inevitably converge on an obvious schema that they just don't want to let the database know about - or develop the Things table that re-implements database schemas poorly. The old table that is clearly several other tables mashed together.
This data is going to be the longest lived and probably most valuable artefact of the company. It is nearly free to put a schema on it, it improves correctness and performance more or less for free. Yet still it isn't possible to get consensus that telling the database the data schema might help.
People don't realise how inefficient JSON blobs are either. I swear there must be AWS databases losing a significant fraction of their IO budget to re-parsing what should be column names with every data point read. I can only hope that postgres devs have rolled their eyes and optimised the case where people don't know how to express tables in SQL.