This is an automated email from the git hooks/post-receive script.
dgoulet pushed a commit to branch main in repository torspec.
commit 1473e9592a84c78833c8d7b422f7ec429c189a59 Author: Mike Perry mikeperry-git@torproject.org AuthorDate: Wed May 17 19:38:16 2023 +0000
Prop#329: Clarity improvements --- proposals/329-traffic-splitting.txt | 119 ++++++++++++++++++++---------------- 1 file changed, 68 insertions(+), 51 deletions(-)
diff --git a/proposals/329-traffic-splitting.txt b/proposals/329-traffic-splitting.txt index 8e7e149..63d7aab 100644 --- a/proposals/329-traffic-splitting.txt +++ b/proposals/329-traffic-splitting.txt @@ -38,26 +38,27 @@ Status: Needs-Revision Tor relay queues, and not with any other bottlenecks (such as intermediate Internet routers), we can avoid this complexity merely by specifying that any paths that are constructed SHOULD NOT share any - relays (except for the exit). This assumption is valid, because non-relay bottlenecks are managed - by TCP of client-to-relay and relay-to-relay OR connections, and not - Tor's circuit-level congestion control. In this way, we can proceed to - use the exact same congestion control as specified in [PROP324], - for each path. + relays (except for the exit). This assumption is valid, because non-relay + bottlenecks are managed by TCP of client-to-relay and relay-to-relay OR + connections, and not Tor's circuit-level congestion control. In this way, + we can proceed to use the exact same congestion control as specified in + [PROP324], for each path.
For this reason, this proposal will focus on protocol specification, and the traffic scheduling algorithms, rather than coupling. Note that the scheduling algorithms are currently in flux, and will be subject to change as we tune them in Shadow, on the live network, and for future - UDP implementation (see [PROP339]). + UDP implementation (see [PROP339]). This proposal will be kept up to + date with the current implementation.
1.2. Divergence from the initial Conflux design
The initial [CONFLUX] paper doesn't provide any indications on how to handle the size of out-of-order cell queue, which we consider a potential dangerous memory DoS vector (see [MEMORY_DOS]). It also used - RTT as the sole heuristic for selecting which circuit to send on, which + RTT as the sole heuristic for selecting which circuit to send on (which may vary depending on the geographical locations of the participant - relays, without considering their actual available circuit capacity + relays), without considering their actual available circuit capacity (which will be available to us via Proposal 324). Additionally, since the publication of [CONFLUX], more modern packet scheduling algorithms have been developed, which aim to reduce out-of-order queue size. @@ -70,8 +71,7 @@ Status: Needs-Revision
1.3. Design Overview
- The following section describes the Conflux design. Each sub-section is - a building block to the multipath design that Conflux proposes. + The following section describes the Conflux design.
The circuit construction is as follows:
@@ -110,10 +110,10 @@ Status: Needs-Revision different scheduling algorithms to determine this.
Initial RTT is measured during circuit linking, as described in - [CONFLUX_HANDSHAKE]. RTT is continually measured using SENDME timing, as - in Proposal 324. This means that during use, the primary circuit and - secondary circuit may switch roles, depending on unrelated network - congestion caused by other Tor clients. + [CONFLUX_HANDSHAKE]. After the initial link, RTT is continually measured + using SENDME timing, as in Proposal 324. This means that during use, + the primary circuit and secondary circuit may switch roles, depending on + unrelated network congestion caused by other Tor clients.
We also support linking onion service circuits together. In this case, only two rendezvous circuits are linked. Each of these RP circuits will @@ -124,11 +124,11 @@ Status: Needs-Revision we have consulted (see [ACKNOWLEDGMENTS]), believe Tor's congestion control from Proposal 324 to be sufficient in this rare case.
- In the algorithms we recommend here, only two circuits will be linked together at a time. - However, implementations - SHOULD support more than two paths, as this has been shown to assist in - traffic analysis resistance[WTF_SPLIT], and will also be useful for - maintaining a desired target RTT, for UDP VoIP applications. + In the algorithms we recommend here, only two circuits will be linked + together at a time. However, implementations SHOULD support more than + two paths, as this has been shown to assist in traffic analysis + resistance[WTF_SPLIT], and will also be useful for maintaining a desired + target RTT, for UDP VoIP applications.
If the number of circuits exceeds the current number of guard relays, guard relays MAY be re-used, but implementations SHOULD use the same @@ -160,11 +160,12 @@ Status: Needs-Revision The "max-num-circ" value indicate the maximum number of rendezvous circuits that are allowed to be linked together.
- We let the service specify the conflux algorithm to use. Some services may - prefer latency, where as some may prefer throughput. However, clients will - also have to be able to override this request, because the high-throughput - algorithms will require more out-of-order queue memory, which may be - infeasible on mobile. + We let the service specify the conflux algorithm to use, when sending data + to the service. Some services may prefer latency, where as some may prefer + throughput. However, clients also have the ability to request their own UX + for data that the service sends, in the LINK handshake below, in part + because the high-throughput algorithms will require more out-of-order queue + memory, which may be infeasible on mobile.
The next section describes how the circuits are linked together.
@@ -263,19 +264,20 @@ Status: Needs-Revision These three relay commands are sent on *each* leg, to allow each endpoint to measure the initial RTT of each leg.
- The client SHOULD abandon and close circuit if the LINKED message takes too long to arrive. - This timeout MUST be no larger than the normal SOCKS/stream timeout in use - for RELAY_BEGIN, but MAY be the Circuit Build Timeout value, instead. - (The C-Tor implementation currently uses Circuit Build Timeout). + The client SHOULD abandon and close circuit if the LINKED message takes too + long to arrive. This timeout MUST be no larger than the normal SOCKS/stream + timeout in use for RELAY_BEGIN, but MAY be the Circuit Build Timeout value, + instead. (The C-Tor implementation currently uses Circuit Build Timeout).
See [SIDE_CHANNELS] for rules for when to reject unexpected handshake cells.
2.2. Linking Circuits from OP to Exit [LINKING_EXIT]
- To link exit circuits, two circuits to the same exit are built. When - each circuit is opened, we ensure that congestion control has been - negotiated. If congestion control negotiation has failed, the circuit - MUST be closed. After this, the linking handshake begins. + To link exit circuits, two circuits to the same exit are built, with + additional restrictions such that they do not share Guard or Middle + relays. When each circuit is opened, we ensure that congestion control + has been negotiated. If congestion control negotiation has failed, the + circuit MUST be closed. After this, the linking handshake begins.
The RTT times between RELAY_CONFLUX_LINK and RELAY_CONFLUX_LINKED are measured by the client, to determine primary vs secondary circuit use, @@ -285,9 +287,11 @@ Status: Needs-Revision
Because of the race between initial data and the RELAY_CONFLUX_LINKED_ACK cell, conditions can arise where an Exit needs to send data before the - slowest circuit delivers this ACK. In these cases, it should prefer the - circuit that has delivered the ACK (which will arrive immediately prior - to any data). + slowest circuit delivers this ACK. In these cases, it should prefer sending + data on the circuit that has delivered the ACK (which will arrive immediately + prior to any data from the client). This circuit will be the lower RTT + circuit anyway, but the code needs to handle the fact that in this case, + there won't yet be an RTT for the second circuit.
2.3. Linking circuits to an onion service [LINKING_SERVICE]
@@ -326,9 +330,9 @@ Status: Needs-Revision In C-Tor, conflux is only used via circuit prebuilding. Pre-built conflux sets are preferred over other pre-built circuits, but if the pre-built pool ends up empty, normal pre-built circuits are used. If those run out, regular - non-conflux circuits are built. Conflux sets are never built on-demand, but - this is strictly an implementation decision, to simplify dealing with the - C-Tor codebase. + non-conflux circuits are built. In other words, in C-Tor, conflux sets are + never built on-demand, but this is strictly an implementation decision, to + simplify dealing with the C-Tor codebase
The consensus parameter 'cfx_max_prebuilt_set' specifies the number of sets to pre-build. @@ -342,13 +346,13 @@ Status: Needs-Revision When a set is launched, legs begin the handshake in the unlinked state. As handshakes complete, finalization is attempted, to create a linked set. On the client, this finalization happens upon receipt of the LINKED cell. - On the exit/service, this finalization happens upon sending the LINKED_ACK. + On the exit/service, this finalization happens upon *sending* the LINKED_ACK.
The initiator of this handshake considers the set fully linked once the RELAY_CONFLUX_LINKED_ACK is sent (roughly upon receipt of the LINKED cell). Because of the potential race between LINKED_ACK, and initial data sent by the client, the receiver of the handshake must consider a leg linked at - the time of sending a LINKED cell. + the time of *sending* a LINKED_ACK cell.
This means that exit legs may not have an RTT measurement, if data on the faster leg beats the LINKED_ACK on the slower leg. The implementation MUST @@ -381,15 +385,15 @@ Status: Needs-Revision legs' maximum last_seq_recv, and a lower last_seq_recv than all current legs last_seq_sent.
- This check is performed on finalization, not the receipt of the cell. This - gives the data additional time to arrive. + This check is performed on finalization, not the receipt of first + handshake cell. This gives the data additional time to arrive.
2.5. Congestion Control Application [CONGESTION_CONTROL]
The SENDMEs for congestion control are performed per-leg. As soon as data arrives, regardless of its ordering, it is counted towards SENDME - delivery. In this way, 'cwnd - package_window' of each leg always - reflects the available data to send on each leg. This is important for + delivery. In this way, 'cwnd - inflight' of each leg always reflects + the available data to send on each leg. This is important for [SCHEDULING].
The Congestion control Stream XON/XOFF can be sent on either leg, and @@ -399,7 +403,19 @@ Status: Needs-Revision of their circuit was blocked. Because conflux can send on the other circuit, which uses a different OR conn, this form of stream blocking has been decoupled from the OR conn status, and only happens when - congestion control has decided that all circuits are blocked. + congestion control has decided that all circuits are blocked (congestion + control becomes blocked when either 'cwnd - inflight <= 0', *or* when + the local OR conn is blocked, so if all local OR conns of a set are + blocked, then the stream will block that way). + + Note also that because congestion control only covers RELAY_COMMAND_DATA + cells, for all algorithms, a special case must be made such that if no + circuit is available to send on due to congestion control blocking, + commands other than RELAY_COMMAN_DATA MUST be sent on the current + circuit, even if the cell scheduler believes that no circuit is available. + Depending on the code structure of Arti, this special case may or may + not be necessary. It arises in C-Tor because nothing can block the + sending of arbitrary non-DATA relay command cells.
2.6. Sequencing [SEQUENCING]
@@ -521,11 +537,12 @@ Status: Needs-Revision RELAY_COMMAND_XON
Currently, this set is the same as the set of cells that have stream ID, - but the property that enforces this is that these cells must be ordered - with respect to all data on the circuit. It is not impossible that future - cells could be invented that don't have stream IDs, but yet must still - arrive in order with respect to circuit data cells. Prop#253 is one - possible example of such a thing (though we won't be implementing that). + but the property that leads to this requirement is not stream usage by + itself, it is that these cells must be ordered with respect to all data + on the circuit. It is not impossible that future relay commands could be + invented that don't have stream IDs, but yet must still arrive in order + with respect to circuit data cells. Prop#253 is one possible example of + such a thing (though we won't be implementing that proposal).
3. Traffic Scheduling [SCHEDULING] @@ -542,7 +559,7 @@ Status: Needs-Revision we create low-memory variants of these that aim to minimize the out-of-order queue size at the receiving endpoint.
- However, see the [TRAFFIC_ANALYSIS] sections of this proposal for + Additionally, see the [TRAFFIC_ANALYSIS] sections of this proposal for important details on how this selection can be changed, to reduce website traffic fingerprinting.