Generally I file them as "devices" for lack of a better description. (IoT stuff)

That discussion is not really about pgsql as such, but more about using a standardised tool which can talk to any database and because it's intended to talk to simple databases, it cripples available queries by disallowing JOIN or subqueries.

I really don't get why people are afraid of pgSQL. MySQL is ok for small jobs but it has a limited range of queries and data types, plus its clustering is badly broken in ways which make it look OK until it falls over unexpectedly (as our devs found out the hard way).

OK, I do get why people are afraid. Yes it uses a bit more memory when you install it, but the days of 64Mb machines are a long way in the past. After having used MySQL for 20 years and then learned to use pgSQL, everything is minor. pgSQL is stricter about some queries but our experience has been "you should have written them better in the first place" - and when fixed they ran faster on the mysql boxes too.

When I converted my Bacula database to pgSQL (50 million entries) it dropped ram requirements by over 60% - 40Gb down to about 7. Queries run 10 times faster and database dumps run 3 times faster. 

MySQL simply doesn't scale. It's good for small, simple databases but GLPI isn't small or simple.

It would also be nice to be able to select pligins by checkbox and enable them as a mass action