Mir

Code review comment for lp:~mir-team/mir/attestable-timestamps-server

Revision history for this message
Tyler Hicks (tyhicks) wrote :

1) Thanks! The copy/paste doc helps a lot but it lacks implementation details. For example, "secret" isn't mentioned anywhere in the document. It does get me closer to understanding the design.

2) I'm confused by the "Content-hub does not need to register itself with Mir" bit since content-hub must share a secret with Mir and then Mir's CookieFactory is constructed from that secret. How does Mir know that the CookieFactory that it constructed was done so with a secret from Content-hub and not some other process?

3) If Mir disconnects a client that submits an invalid cookie, what prevents the client from reconnecting and submitting another invalid cookie?
  - Documenting that any attestation failure should be fatal to the client would be a good thing as long as a reconnecting is expensive.
  - IIUC, assert_timestamp() is in the address space of the client, correct? If so, throwing an exception on failure would do no good because a malicious client could link against a modified library that doesn't throw an exception.

review: Needs Information

« Back to merge proposal