And really : ading it in core can broke completely Expression Manager …

So we are basically sitting on a vulcanoe? because if understand you right, it is extremely dangerous to make any changes, additions, etc. to EM?

holch wrote: …it is extremely dangerous to make any changes, additions, etc. to EM?

Some specific update : yes … (maybe not extremely …)

But there are another think : currently all EM data are saved in session, we already have a big session … really big session …

I really think .question (and .help if we add it) must not be in session (i really think 90% of EM session muts not be in session too)
before adding big part in EM, i prefer to update system to have a ligthten session …

holch wrote: ...
- Subquestion relevance but now answer option relevance. We understand this difference, but for many users subquestions are basically the same as answer options.

Just for the record: This is very true.

As a user, I would have already wished more than once for answer option relevance and certainly do not understand why it is not there (although subquestion relevance is).

(Of course I do understand that LS is open source and therefore awesome that it exists at all the way it does atm. So, no harm meant, I don't want to come across as too critical. You LS developer guys ROCK!!)

But there are another think : currently all EM data are saved in session, we already have a big session … really big session …

Totally agree! The session should not grow any further. Actually we need to reduce the session as much as possible.

I really think .question (and .help if we add it) must not be in session (i really think 90% of EM session muts not be in session too)
before adding big part in EM, i prefer to update system to have a lighten session …

What implications would this have, if we get this from the database? Would this slow things down considerably?

As a user, I would have already wished more than once for answer option relevance and certainly do not understand why it is not there (although subquestion relevance is).

I guess we all wished for answer option relevance already various times. But then, to be honest, subquestion relevance hasn't always been there. Not long ago, we did not even have subquestion relevance.
Problem is, that answer options and subquestions are two totally different concepts in the code of Limesurvey. This is not so obvious to users and thus it is hard to understand why one exists and the other doesn't. But I guess there would be quite some work involved to get this implemented. And from what I understand, LS carries around a lot of old code, which makes it difficult to maintain.

holch wrote: What implications would this have, if we get this from the database? Would this slow things down considerably?

I don't really know

For javascript : I think EM find needed knowVars and put it in page (a javascript var). Get a single information like question and help, is really cheap for DB
For PHP : we ask only if we need it : really cheap too.