Re: [hibernate-dev] Second-level cache and Inheritance

Hi Vlad, I think that might be quite tricky to implement, as it might not always be clear into which region we'd need to lookup from, or invalidate, not least locks might need to cross multiple regions. Think for example when a type A is removed with ID=1, and then a type B with the same ID is being re-inserted as a replacement, but "upgrading" the type, e.g. B extends A so it's similar to the original value but adds some more fields. Not that I'd ever recommend doing such weird things, but we obviously need to be correct to deal with such scenarios. If you then also throw in the fact that multiple such regions which need to be treated correctly from a concurrency point of view, as a combination of multiple different possible concurrency strategies, then the code complexity explodes even further and there are many details to take into account.. Assuming even that someone was up to the challenge to implement this (and get it right?!), you'd pay a significant price in terms of runtime performance due to this complexity. -- Sanne On 18 July 2016 at 11:10, Vlad Mihalcea <mihalcea.vlad(a)gmail.com&gt; wrote: