Comment 23 for bug 1449212

so I think the mostly unit-test fallout is going to be an issue for back-porting. The commit message is great, but it doesn't really go into detail about why moving the env cache to the subdir was required.

I think I understand now that in creating a copy of the subrequest before sending it down into the app the cached values weren't getting set on the req that originally came *into* the app - which caused some things to go south because of how crazy all that env caching get_info stuff is. Probably should be fixed - but not 100% sure it needs to be part of this fix and backport? Honestly the whole copy trick seems like a lot of effort to go through to avoid the swift.authorize pop? Why can't we just do like what we do with cross-account-copy requests and just leave the authorize callback in place?

https://gist.github.com/clayg/11e4e9c0c3f652d82e0f

^ this mostly just steals sam's hardwork and tries to avoid the info cache change that caused all the test fallout.

Also I'm not 100% sure on the precedence of container and account tempurl keys? Maybe it was pre-existing but I don't think it had the same effect. Now if a tempurl matches the container key it'll get scoped - but if that key happens to collide with the account my cross container segment object downloads don't authorize - and there's nothing really to tell me whats going on? Would it be better to scope to the larger matching scope?

Also, I think we need functests to cover this issue.

Also, isn't there a separate issue with handing out a PUT tempurl that we need to address with stripping manifest headers - or did I miss where we decided that's a non-issue?