Bertrand Delacretaz wrote > On Thu, Mar 3, 2016 at 8:26 AM, Carsten Ziegeler <cziege...@apache.org> wrote: >> ...I was thinking of rewriting the whole script resolution mechanism to >> have a better caching mechanism. Maybe if we do one of these things we >> look into the other as well :).. > > I think so - including some form of "context key" in the cache keys > makes multiple resolution contexts possible, so rewriting with this > new use case in mind should help. > Today, the cache is totally sub optimal as it does not take the available scripts into account, but it uses the input from the request (resource types, extensions, selectors) as a key. And therefore the cache blows up with extensive selector usage. We need to change that and build keys based on the available scripts (and therefore selectors that are handled). This reduces the key space for cache keys and makes the cache way more efficient, regardless if users use selectors to pass parameters or not. I think this shouldn't be too hard to do, but it requires some changes and at the same time we should also rethink other things like handling the resource type hierarchy etc. I was planning to write RTs about those two topics in the near future, still need some time to properly write it down :(
Regards Carsten -- Carsten Ziegeler Adobe Research Switzerland cziege...@apache.org
