Closed hfaulds closed 9 years ago
The rationale for this PR is that we don't use a lot of the metadata we fetch but we cache it. In some mlses this cached metadata is so large it's causing us problems.
I have reservations about this approach - it adds a decent amount of complexity and some really weird edge cases when coupled with any sort of caching. I'd prefer to break the metadata into cacheable chunks by resource or something like that.
Needs some tests but is safe to merge in terms of not affecting any existing functionality because it's all behind an option.