Filename: 158-microdescriptors.txt
Title: Clients download consensus + microdescriptors
Author: Roger Dingledine
Created: 17-Jan-2009
Status: Closed
Implemented-In: 0.2.3.1-alpha

0. History

  15 May 2009: Substantially revised based on discussions on or-dev
  from late January.  Removed the notion of voting on how to choose
  microdescriptors; made it just a function of the consensus method.
  (This lets us avoid the possibility of "desynchronization.")
  Added suggestion to use a new consensus flavor.  Specified use of
  SHA256 for new hashes. -nickm

  15 June 2009: Cleaned up based on comments from Roger. -nickm

1. Overview

  This proposal replaces section 3.2 of proposal 141, which was
  called "Fetching descriptors on demand". Rather than modifying the
  circuit-building protocol to fetch a server descriptor inline at each
  circuit extend, we instead put all of the information that clients need
  either into the consensus itself, or into a new set of data about each
  relay called a microdescriptor.

  Descriptor elements that are small and frequently changing should go
  in the consensus itself, and descriptor elements that are small and
  relatively static should go in the microdescriptor. If we ever end up
  with descriptor elements that aren't small yet clients need to know
  them, we'll need to resume considering some design like the one in
  proposal 141.

  Note also that any descriptor element which clients need to use to
  decide which servers to fetch info about, or which servers to fetch
  info from, needs to stay in the consensus.

2. Motivation

  See
  http://archives.seul.org/or/dev/Nov-2008/msg00000.html and
  http://archives.seul.org/or/dev/Nov-2008/msg00001.html and especially
  http://archives.seul.org/or/dev/Nov-2008/msg00007.html
  for a discussion of the options and why this is currently the best
  approach.

3. Design

  There are three pieces to the proposal. First, authorities will list in
  their votes (and thus in the consensus) the expected hash of
  microdescriptor for each relay. Second, authorities will serve
  microdescriptors, directory mirrors will cache and serve
  them. Third, clients will ask for them and cache them.

3.1. Consensus changes

  If the authorities choose a consensus method of a given version or
  later, a microdescriptor format is implicit in that version.
  A microdescriptor should in every case be a pure function of the
  router descriptor and the consensus method.

  In votes, we need to include the hash of each expected microdescriptor
  in the routerstatus section. I suggest a new "m" line for each stanza,
  with the base64 of the SHA256 hash of the router's microdescriptor.

  For every consensus method that an authority supports, it includes a
  separate "m" line in each router section of its vote, containing:
    "m" SP methods 1*(SP AlgorithmName "=" digest) NL
  where methods is a comma-separated list of the consensus methods
  that the authority believes will produce "digest".

  (As with base64 encoding of SHA1 hashes in consensuses, let's
  omit the trailing =s)

  The consensus microdescriptor-elements and "m" lines are then computed
  as described in Section 3.1.2 below.

  (This means we need a new consensus-method that knows
  how to compute the microdescriptor-elements and add "m" lines.)

  The microdescriptor consensus uses the directory-signature format from
  proposal 162, with the "sha256" algorithm.


3.1.1. Descriptor elements to include for now

  In the first version, the microdescriptor should contain the
  onion-key element, and the family element from the router descriptor,
  and the exit policy summary as currently specified in dir-spec.txt.

3.1.2. Computing consensus for microdescriptor-elements and "m" lines

  When we are generating a consensus, we use whichever m line
  unambiguously corresponds to the descriptor digest that will be
  included in the consensus.

  (If different votes have different microdescriptor digests for a
  single <descriptor-digest, consensus-method> pair, then at least one
  of the authorities is broken.  If this happens, the consensus should
  contain whichever microdescriptor digest is most common.  If there is
  no winner, we break ties in the favor of the lexically earliest.
  Either way, we should log a warning: there is definitely a bug.)

  The "m" lines in a consensus contain only the digest, not a list of
  consensus methods.

3.1.3. A new flavor of consensus

  Rather than inserting "m" lines in the current consensus format,
  they should be included in a new consensus flavor (see proposal
  162).

  This flavor can safely omit descriptor digests.

  When we implement this voting method, we can remove the exit policy
  summary from the current "ns" flavor of consensus, since no current
  clients use them, and they take up about 5% of the compressed
  consensus.

  This new consensus flavor should be signed with the sha256 signature
  format as documented in proposal 162.

3.2. Directory mirrors fetch, cache, and serve microdescriptors

  Directory mirrors should fetch, catch, and serve each microdescriptor
  from the authorities.  (They need to continue to serve normal relay
  descriptors too, to handle old clients.)

  The microdescriptors with base64 hashes <D1>,<D2>,<D3> should be
  available at:
    http://<hostname>/tor/micro/d/<D1>-<D2>-<D3>.z
  (We use base64 for size and for consistency with the consensus
  format. We use -s instead of +s to separate these items, since
  the + character is used in base64 encoding.)

  All the microdescriptors from the current consensus should also be
  available at:
    http://<hostname>/tor/micro/all.z
  so a client that's bootstrapping doesn't need to send a 70KB URL just
  to name every microdescriptor it's looking for.

  Microdescriptors have no header or footer.
  The hash of the microdescriptor is simply the hash of the concatenated
  elements.

  Directory mirrors should check to make sure that the microdescriptors
  they're about to serve match the right hashes (either the hashes from
  the fetch URL or the hashes from the consensus, respectively).

  We will probably want to consider some sort of smart data structure to
  be able to quickly convert microdescriptor hashes into the appropriate
  microdescriptor. Clients will want this anyway when they load their
  microdescriptor cache and want to match it up with the consensus to
  see what's missing.

3.3. Clients fetch them and cache them

  When a client gets a new consensus, it looks to see if there are any
  microdescriptors it needs to learn. If it needs to learn more than
  some threshold of the microdescriptors (half?), it requests 'all',
  else it requests only the missing ones.  Clients MAY try to
  determine whether the upload bandwidth for listing the
  microdescriptors they want is more or less than the download
  bandwidth for the microdescriptors they do not want.

  Clients maintain a cache of microdescriptors along with metadata like
  when it was last referenced by a consensus, and which identity key
  it corresponds to.  They keep a microdescriptor
  until it hasn't been mentioned in any consensus for a week. Future
  clients might cache them for longer or shorter times.

3.3.1. Information leaks from clients

  If a client asks you for a set of microdescs, then you know she didn't
  have them cached before. How much does that leak? What about when
  we're all using our entry guards as directory guards, and we've seen
  that user make a bunch of circuits already?

  Fetching "all" when you need at least half is a good first order fix,
  but might not be all there is to it.

  Another future option would be to fetch some of the microdescriptors
  anonymously (via a Tor circuit).

  Another crazy option (Roger's phrasing) is to do decoy fetches as
  well.

4. Transition and deployment

  Phase one, the directory authorities should start voting on
  microdescriptors, and putting them in the consensus.

  Phase two, directory mirrors should learn how to serve them, and learn
  how to read the consensus to find out what they should be serving.

  Phase three, clients should start fetching and caching them instead
  of normal descriptors.