Major Description
Pierre-Arnaud have had a great idea ! While I was doing some profiling on a lookup operation, it shown that cloning the entry we were looking for takes around 40% of all the CPU.
His idea is just to wrap this entry and compute the modification only when needed. If we think about the lookup operation, which is a simple search where we just ask for an entry given its DN, plus an optional list of attributes to return, it would be easy to process the returned attributes when constructing the response without modifying the initial entry.
That could save a lost of cycle in the common cases where the user just request the users attributes, as we wil just have to filter the operational attributes.
This is a great idea, and I want to have a trace of it.