The Register has an interesting debate going on:
https://www.theregister.com/2023/03/08/great_graph_debate_wednesday/?td=rt-3a
There is also a poll on the site.
It’s been interesting to watch SQL return from the dead over the past few years…
I’ve always regarded NoSql in general as something that I would at most use in conjunction with a relational DB for a subset of functionality such as storage of just the unstructured data maybe. In practice I haven’t encountered a situation where the architecture of a SQL DB is awkward or “feels” wrong but I’m sure there are applications out there where using Mongo or Cassandra or something for a portion of the DB side of a project could be helpful.
But apart from such simple applications, I feel that NoSql makes it way too easy to design yourself into a box canyon that you can’t get out of. You can make a NoSql “schema” that is superior / faster than a relational DB for a specific use case / usage pattern, but then when inevitably you need to change it, it’s sometimes very difficult to refactor. Or at least that was my conclusion the last time I did a deep dive on MongoDB. Just simply changes in usage patterns can break one of these “perfect” designs when simply adding or adjusting an index in a relational DB would solve the problem.
1 Like