Code review comment for lp:~mir-team/mir/handle-mis-shaped-cookies

I'm not sure what we're /really/ trying to support here. (Beyond avoiding a crash.)

If there are "assumptions about cookies" embedded in the wrong code (i.e. code that isn't reconfigurable with the cookie authority), then the problem is the code is in the wrong place!

Almost every time we call cookie->serialize() we have code than immediately copies the result to a cookie::Blob. If cookie::Cookie had a method to populate a blob directly, then that method could be configured by a custom cookie implementation.

Would that address the problem more effectively?