prop. 160 big update

This commit is contained in:
zzz
2025-04-23 12:43:55 -04:00
parent 27c7158784
commit 12ce4cd852

View File

@ -5,9 +5,9 @@ UDP Trackers
:author: zzz
:created: 2022-01-03
:thread: http://zzz.i2p/topics/1634
:lastupdated: 2022-12-19
:lastupdated: 2025-04-23
:status: Open
:target: 0.9.58
:target: 0.9.68
.. contents::
@ -18,6 +18,36 @@ Overview
This proposal is for implemention of UDP trackers in I2P.
Change History
--------------
A preliminary proposal for UDP trackers in I2P was posted on our bittorrent spec page [SPEC]_
in May 2014; this predated our formal proposal process, and it was never implemented.
This proposal was created in early 2022 and simplifies the 2014 version.
As this proposal relies on repliable datagrams, it was put on hold once we
started working on the Datagram2 proposal [Prop163]_ in early 2023.
That proposal was approved in April 2025.
The 2023 version of this proposal specified two modes, "compatibility" and "fast".
Further analysis revealed that the fast mode would be insecure, and would also
be inefficient for clients with a large number of torrents.
Further, BiglyBT indicated a preference for compatibility mode.
This mode will be easier to implement for any tracker or client supporting
standard [BEP15]_.
While the compatibility mode is more complex to implement from-scratch
on the client size, we do have preliminary code for it started in 2023.
Therefore, the current version here is further simplified to remove fast mode,
and remove the term "compatibility". The current version switches to
the new Datagram2 format, and adds references to the UDP announce extenstion
protocol [BEP41]_.
Also, a connection ID lifetime field is added to the connect response,
to extend the efficiency gains of this protocol.
Motivation
==========
@ -27,10 +57,6 @@ we need to make trackers and announces more efficient so that trackers are not o
Bittorrent proposed UDP trackers in BEP 15 [BEP15]_ in 2008, and the vast majority
of trackers on clearnet are now UDP-only.
A preliminary proposal for UDP trackers in I2P was posted on our bittorrent spec page [SPEC]_
in May 2014; this predated our formal proposal process, and it was never implemented.
This proposal simplifies the 2014 version.
It is difficult to calculate the bandwidth savings of datagrams vs. streaming protocol.
A repliable request is about the same size as a streaming SYN, but the payload
is about 500 bytes smaller because the HTTP GET has a huge 600 byte
@ -41,19 +67,25 @@ for a tracker's outbound traffic.
Additionally, there should be implementation-specific memory reductions,
as datagrams require much less in-memory state than a streaming connection.
Post-Quantum encryption and signatures as envisioned in [Prop169]_ will substantially
increase the overhead of encrypted and signed structures, including destinations,
leasesets, streaming SYN and SYN ACK. It is important to minimize this
overhead where possible before PQ crypto is adopted in I2P.
Design
============
This proposal uses both repliable and raw datagrams,
This proposal uses both repliable datagram2 and raw datagrams,
as defined in [DATAGRAMS]_.
Datagram2 is a new variant of repliable datagrams,
defined in Proposal 163 [Prop163]_, that adds replay resistance.
BEP 15
-------
The message flow in [BEP15]_ is as follows:
For reference, the message flow defined in [BEP15]_ is as follows:
.. raw:: html
@ -69,23 +101,17 @@ Client Tracker
The connect phase is required to prevent IP address spoofing.
The tracker returns a connection ID that the client uses in subsequent announces.
This connection ID expires in one minute at the client, and in two minutes at the tracker.
This is not necessary in I2P because of repliable datagrams.
This connection ID expires by default in one minute at the client, and in two minutes at the tracker.
We propose two mechanisms for I2P, compatibility mode and fast mode.
Compatibility Mode
-------------------------
In compatibility mode, we keep the same message flow as BEP 15,
I2P will use the same message flow as BEP 15,
for ease of adoption in existing UDP-capable client code bases:
for efficiency, and for security reasons discussed below:
.. raw:: html
{% highlight %}
Client Tracker
Connect Req. -------------> (Repliable)
Connect Req. -------------> (Repliable Datagram2)
<-------------- Connect Resp. (Raw)
Announce Req. -------------> (Raw)
<-------------- Announce Resp. (Raw)
@ -94,41 +120,81 @@ Client Tracker
...
{% endhighlight %}
This mode is also useful if the client plans to send multiple announces
within one minute to a single tracker, as only the connect
message must be repliable.
This potentially provides a large bandwidth savings over
streaming (TCP) announces.
While the Datagram2 is about the same size as a streaming SYN,
the raw response is much smaller than the streaming SYN ACK.
All subsequent requests/responses are raw.
I2P Fast Mode
Non-Java Tracker/Client support
-------------------------------
For an integrated application (router and client in one process, for example i2psnark and the ZzzOT Java plugin),
it should be straightforward to implement and route the streaming and datagram traffic separately.
ZzzOT and i2psnark are expected to be the first tracker and client to implement this proposal.
For an external tracker application that currently uses an HTTP server tunnel to receive
announce requests, the implementation could be quite difficult.
A specialized tunnel could be developed to translate datagrams to local HTTP requests/responses.
Or, a specialized tunnel that handles both HTTP requests and datagrams could be designed
that would forward the datagrams to the external process.
These design decisions will depend heavily on the specific router and tracker implementations,
and are outside the scope of this proposal.
External SAM-based torrent clients such as qbittorrent and other libtorrent-based clients
would require SAM v3.3 [SAMv3]_ which is not supported by i2pd.
This is also required for DHT support, and is complex enough that no known
SAM torrent client has implemented it.
Connection Lifetime
-------------------
[BEP15]_ specifies that the connection ID expires in one minute at the client, and in two minutes at the tracker.
It is not configurable.
That limits the potential efficiency gains, unless
clients batched announces to do all of them within a one-minute window.
i2psnark does not currently batch announces; it spreads them out, to avoid bursts of traffic.
Power users are reported to be running thousands of torrents at once,
and bursting that many announces into one minute is not realistic.
Here, we propose to extend the connect response to add an optional connection lifetime field.
The default, if not present, is one minute. Otherwise, the lifetime specified
in seconds, shall be used by the client, and the tracker will maintain the
connection ID for one minute more.
Compatibility with BEP 15
-------------------------
In fast mode, we omit the connect phase, as it is not required to prevent address spoofing.
This significantly simplifies the client-side implementation.
This design maintains compatibility with [BEP15]_ as much as possible
to limit changes required in existing clients and trackers.
.. raw:: html
The only required change is the format of peer info in the announce response.
The addition of the lifetime field in the connect response is not required
but is strongly recommended for efficiency, as explained above.
{% highlight %}
Client Tracker
Announce Req. -------------> (Repliable)
<-------------- Announce Resp. (Raw)
{% endhighlight %}
This mode omits a round-trip, but requires every announce request to be repliable.
Security Analysis
------------------
The primary goal of an announce protocol is to impose a cost on address spoofing.
In compatibility mode, the client must actually exist and bundle a real leaseset.
An important goal of a UDP announce protocol is to prevent address spoofing.
The client must actually exist and bundle a real leaseset.
It must have inbound tunnels to receive the Connect Response.
These tunnels could be zero-hop and built instantly, but that would
expose the creator.
This protocol accomplishes that goal.
However, in fast mode, the destination and leaseset could be fake.
Multiple fake destinations and leasesets can be rapidly generated without
actually building tunnels. The Announce Request messages could then be sent out
any tunnel.
Issues
------
- This proposal does not support blinded destinations,
but may be extended to do so. See below.
@ -136,15 +202,55 @@ any tunnel.
Specification
=============
Repliable datagrams use I2CP protocol 17; raw datagrams use I2CP protocol 18.
Requests may be repliable or raw. Responses are always raw.
Protocols and Ports
-------------------
Repliable Datagram2 uses I2CP protocol 19; raw datagrams use I2CP protocol 18.
Requests may be Datagram2 or raw. Responses are always raw.
The older repliable datagram ("Datagram1") format using I2CP protocol 17
must NOT be used for requests or replies; these must be dropped if received
on the request/reply ports. Note that Datagram1 protocol 17
is still used for the DHT protocol.
Requests use the I2CP "to port" from the announce URL; see below.
The request "from port" is chosen by the client, but should be nonzero,
and a different port from those used by DHT, so that responses
may be easily classified.
Trackers should reject requests received on the wrong port.
Responses use the I2CP "to port" from the request.
The request "from port" is the "to port" from the request.
Announce URL
------------
The announce URL format is not specified in [BEP15]_,
but as in clearnet, UDP announce URLs are of the form "udp://host:port/path".
The path is ignored and may be empty, but is typically "/announce" on clearnet.
The :port part should always be present, however,
if the ":port" part is omitted, use a default I2CP port of 6969,
as that is the common port on clearnet.
There may also be cgi parameters &a=b&c=d appended,
those may be processed and provided in the announce request, see [BEP41]_.
If there are no parameters or path, the trailing / may also be omitted,
as implied in [BEP41]_.
Datagram Formats
----------------
All values are send in network byte order (big endian).
Do not expect packets to be exactly of a certain size.
Future extensions could increase the size of packets.
Connect Request
-----------------
```````````````
Client to tracker.
16 bytes. Must be repliable. Same as in [BEP15]_.
16 bytes. Must be repliable Datagram2. Same as in [BEP15]_. No changes.
.. raw:: html
@ -159,10 +265,10 @@ Offset Size Name Value
Connect Response
-----------------
````````````````
Tracker to client.
16 bytes. Must be raw. Same as in [BEP15]_.
16 or 18 bytes. Must be raw. Same as in [BEP15]_ except as noted below.
.. raw:: html
@ -172,26 +278,27 @@ Offset Size Name Value
0 32-bit integer action 0 // connect
4 32-bit integer transaction_id
8 64-bit integer connection_id
16 16-bit integer lifetime optional // Change from BEP 15
{% endhighlight %}
The response MUST be sent to the I2CP "to port" that was received as the request "from port".
The lifetime field is optional and indicates the connection_id client lifetime in seconds.
The default is 60, and the minimum if specified is 60.
The maximum is 65535 or about 18 hours.
The tracker should maintain the connection_id for 60 seconds more than the client lifetime.
Announce Request
-----------------
````````````````
Client to tracker.
98 bytes. Same as in [BEP15]_ except as noted below.
98 bytes minimum. Must be raw. Same as in [BEP15]_ except as noted below.
If preceded by a connect request/response, must be raw,
with the connection_id received in the connect response.
The connection_id is as received in the connect response.
If NOT preceded by a connect request/response, must be repliable,
and the connection_id is ignored.
.. raw:: html
@ -210,30 +317,25 @@ Offset Size Name Value
88 32-bit integer key
92 32-bit integer num_want -1 // default
96 16-bit integer port
98 TBD additional data TBD
98 varies options optional // As specified in BEP 41
{% endhighlight %}
Changes from [BEP15]_:
- connection_id is ignored if repliable
- IP address is ignored
- key is ignored
- port is probably ignored
- Explicitly indidate that the protocol is extensible,
with possible additional data starting at port 98.
- The options section, if present, is as defined in [BEP41]_
The response MUST be sent to the I2CP "to port" that was received as the request "from port".
Do not use the port from the announce request.
Announce Response
-----------------
`````````````````
Tracker to client.
20+ bytes. Must be raw. Same as in [BEP15]_ except as noted below.
20 bytes minimum. Must be raw. Same as in [BEP15]_ except as noted below.
@ -246,74 +348,67 @@ Offset Size Name Value
8 32-bit integer interval
12 32-bit integer leechers
16 32-bit integer seeders
20 16-bit integer count of hashes to follow
22 32 * n 32-byte hash binary hashes
...
22 + 32 * n TBD additional data TBD
20 32 * n 32-byte hash binary hashes // Change from BEP 15
... // Change from BEP 15
{% endhighlight %}
Changes from [BEP15]_:
- Add a hash count before the hashes, so that the response format
is extensible with additional data after the hashes.
- Instead of 6-byte IPv4+port or 18-byte IPv6+port, we return
a multiple of 32-byte "compact responses" with the SHA-256 binary peer hashes.
As with TCP compact responses, we do not include a port.
- Explicitly indidate that the protocol is extensible,
with possible additional data starting after the hashes
The response MUST be sent to the I2CP "to port" that was received as the request "from port".
Do not use the port from the announce request.
I2P datagrams have a very large maximum size of about 16 KB;
I2P datagrams have a very large maximum size of about 64 KB;
however, for reliable delivery, datagrams larger than 4 KB should be avoided.
For bandwidth efficiency, trackers should probably limit the maximum peers
to about 50.
to about 50, which corresponds to about a 1600 byte packet before overhead
at various layers, and should be within a two-tunnel-message payload limit
after fragmentation.
As in BEP 15, there is no count included of the number of peer addresses
(IP/port for BEP 15, hashes here) to follow.
While not contemplated in BEP 15, an end-of-peers marker
of all zeros could be defined to indicate that the peer info is complete
and some extension data follows.
So that extension is possible in the future, clients should ignore
a 32-byte all-zeros hash, and any data that follows.
Trackers should reject announces from an all-zeros hash,
although that hash is already banned by Java routers.
Scrape
----------
``````
Scrape request/response from [BEP15]_ is not required by this proposal,
but may be implemented if desired, no changes required.
The scrape request is always repliable (unless there is a previous connect request/response)
and the scrape response is always raw.
The client must acquire a connection ID first.
The scrape request and response are always raw.
Error Response
------------------
``````````````
Error response from [BEP15]_ is not required by this proposal,
but may be implemented if desired, no changes required.
The error response is always raw.
Tracker to client.
8 bytes minimum (if the message is empty).
Must be raw. Same as in [BEP15]_. No changes.
.. raw:: html
Announce URL
------------
{% highlight %}
As in clearnet, UDP announce URLs are of the form "udp://host:port/path".
The path is ignored and may be empty.
If the ":port" part is omitted, use an I2CP port of 0.
Offset Size Name Value
0 32-bit integer action 3 // error
4 32-bit integer transaction_id
8 string message
{% endhighlight %}
Issues
=======
- Repliable datagrams do not support offline signatures.
That requires a separate proposal.
- This proposal does not support blinded destinations,
but may be extended to do so. See below.
- This proposal offers two modes at the client's option.
An existing clearnet tracker such as "opentracker" would require more modifications
to support the fast mode. There is no way in the announce URL to indicate
support for only one mode.
- Compatibility mode may not be necessary, pending feedback from BiglyBT and
other developers. However, it would still save a lot of bandwidth
if it is used for several announces within a minute.
Repliable announces are about 450 bytes larger than raw announces.
Extensions
=============
@ -321,13 +416,16 @@ Extensions
Extension bits or a version field are not included.
Clients and trackers should not assume packets to be of a certain size.
This way, additional fields can be added without breaking compatibility.
The extensions format defined in [BEP41]_ is recommended if required.
The announce response is modified to include a count of peer hashes,
so that the response may be easily extended with additional information.
The connect response is modified to add an optional connection ID lifetime.
If blinded destination support is required, we can either add the
blinded 35-byte address to the end of the announce request, or define a new blinded announce request message.
The set of blinded 35-byte peer addresses could be added to the end of the announce reply.
blinded 35-byte address to the end of the announce request,
or request blinded hashes in the responses,
using the [BEP41]_ format (paramters TBD).
The set of blinded 35-byte peer addresses could be added to the end of the announce reply,
after an all-zeros 32-byte hash.
@ -340,17 +438,13 @@ Clients
For a given tracker hostname, a client should prefer UDP over HTTP URLs,
and should not announce to both.
Clients wihout existing BEP 15 support should implement
fast mode only, as it is much simpler.
Clients with existing BEP 15 support should require only small modifications.
Evaluate both fast and compatibility modes and choose
whatever is best for the existing code base.
If a client support DHT or other datagram protocols, it should probably
select a different port as the request "from port" so that the replies
come back to that port and are not mixed up with DHT messages.
The client only receives raw datagrams as replies.
Trackers will never send a repliable datagram to the client.
Trackers will never send a repliable datagram2 to the client.
Clients with a default list of opentrackers should update the list to
add UDP URLs after the known opentrackers are known to support UDP.
@ -360,26 +454,15 @@ Retransmissions, if implemented, should use an initial timeout
of at least 15 seconds, and double the timeout for each retransmission
(exponential backoff).
Clients must back off after receiving an error response.
Trackers
---------
Trackers must implement both compatibility mode and fast mode.
Trackers with existing BEP 15 support should require only small modifications.
This proposal differs from the 2014 proposal, in that the tracker
must support reception of repliable and raw datagrams on the same port.
For an integrated application (router and client in one process, for example the ZzzOT Java plugin),
it should be straightforward to implement and route the streaming and datagram traffic separately.
For an external tracker application that currently uses an HTTP server tunnel to receive
announce requests, the implementation could be quite difficult.
A specialized tunnel could be developed to translate datagrams to local HTTP requests/responses.
Or, a specialized tunnel that handles both HTTP requests and datagrams could be designed
that would forward the datagrams to the external process.
These design decisions will depend heavily on the specific router and tracker implementations,
and are outside the scope of this proposal.
must support reception of repliable datagram2 and raw datagrams on the same port.
@ -388,7 +471,7 @@ Migration
Existing clients do not support UDP announce URLs and ignore them.
Existing trackers do not support reception of repliable or raw datagrams, they will be dropped.
Existing trackers do not support reception of repliable datagram2 or raw datagrams, they will be dropped.
This proposal is completely optional. Neither clients nor trackers are required to implement it at any time.
@ -411,8 +494,20 @@ References
.. [BEP15]
http://www.bittorrent.org/beps/bep_0015.html
.. [BEP41]
http://www.bittorrent.org/beps/bep_0041.html
.. [DATAGRAMS]
{{ spec_url('datagrams') }}
.. [Prop163]
{{ proposal_url('163') }}
.. [Prop169]
{{ proposal_url('169') }}
.. [SAMv3]
{{ site_url('docs/api/samv3') }}
.. [SPEC]
{{ site_url('docs/applications/bittorrent', True) }}