> Is easy to underestimate them.
VERY true.
> Databases are not easy to handle and operate.
I think instead are the easiest of all available options. With a SQL database you have at least available some kind of GUI to build the schema and do query and stuff.
And have see not-very-technical people building databases just fine.
Try the same with Cassandra "Hey, secretary, please build a Cassandra database for the contacts, please!".
Yep, not easy.
------
You only get into trouble when:
- You schema is super-bad. But is the same with NoSql
- You don't create that 1-2 indexes per-table (and some Guis/RDBMS auto-create them for the ID and surrogate keys, that is almost enough)
- You have a very large database. Or, you have a bad schema and is leaking as a very large resultset (or SQL query) that is overcomplicated everything.
- You forgot you have a powerfull RDBMS and insist to use it below the capabilities of Acces, yet, have proud in learning C++. Because C++ is powerful. But why dedicate a few hours learning the basic of Sql??
Most people have very easy problems to solve, and in contrast with the nightmare that is debug a NoSql store that badly try to re-create what a RDBMS have and you know some data is lost but good look finding it...
---
Where the situation get out of proportion is with the startup-mindset of "scalability". A problem that few have, even startups.