proxying to loggerhead won't work for private branches

[Martin Pool]

https://code.launchpad.net/~spiv/launchpad/bmp-inline-diffs/+merge/66634 and bug 813349 and friends take the approach of proxying client requests through the main app domain to loggerhead.

This probably isn't going to work for private branches, because (iiuc) loggerhead uses a separate authentication cookie, so the client is going to send the cookie relevant to say code.l.n and loggerhead won't know what to do with it.

One solution would be to teach loggerhead to validate the cookie by calling across to lp.

[Martin Pool (mbp)]

I can confirm that with that code active on qas, you do indeed get a reasonably clean (but cryptic) little 'Error' when you try to expand a revision from a private branch. I think this is acceptable for a beta.

flacoste thought perhaps loggerhead used the same session cookies, but this seems to show that it does not.

[Michael Hudson-Doyle (mwhudson)]

Given that loggerhead calls translatePath(user, branch) to access a branch anyway, it could call translatePathCookie(session_cookie, branch) instead. This seems most sensible, according to a small number of developers on irc :)