I would like to announce torrents to my opentracker using an Elixir program. I have a test qbittorrent and opentracker running and I’ve identified how qbittorrent sends it’s announces to opentracker.

METHOD: GET
URL
http://localhost:8000/announce?info_hash=%ac%c3%b2%e43%d7%c7GZ%bbYA%b5h%1c%b7%a1%ea%26%e2&peer_id=-qB5020-r3FSX0qNU6Oo&port=4993&uploaded=0&downloaded=0&left=0&corrupt=0&key=F7F879A3&event=started&numwant=200&compact=1&no_peer_id=1&supportcrypto=1&redundant=0
HEADERS
Accept-Encoding:
gzip
Connection:
close
Host:
localhost:8000
User-Agent:
qBittorrent/5.0.2

I would like to build this same GET request into my own code. My confusion comes from the percent encoded info_hash. Neither URI.encode/2 nor URI.encode_www_form/1 encode the URI the same way that qbittorrent does it.

Let me show you what I’ve tried so far.

# We start with the `Info hash v1` of the torrent, as copied from qBittorrent.
info_hash_v1 = "acc3b2e433d7c7475abb5941b5681cb7a1ea26e2"

# Next we decode the hexadecimal representation into binary.
binary = Base.decode16!(info_hash_v1, case: :lower)

I’m not sure what comes next. Digging through Elixir URI source code, I can see the method of percent encoding that creates a binary in almost the right format.

The problem here is that the hex/1 function only outputs 16 possible values, A-F, 0-9. There is no lowercase! The way qbittorrent is doing their percent encoding, they have lowercase too-- a-f, A-F, 0-9.

I read qbittorrent source code to see how they’re percent encoding, but I couldn’t find the code that does that. I found some info_hash references though. I can barely read C++, but I think the url encoding might be abstracted away in a request library.

I also looked through transmission-qt code. That code is greek to me, but I was able to find their percent encoder implementation.

Seems similar to what I read in Elixir URI source code related to unreserved and unescaped characters. RFC 3986 - Uniform Resource Identifier (URI): Generic Syntax

I bounced some ideas off of ChatGPT, borrowed code from Elixir URI, and I got a bep3_encode/1 function put together.

  @doc """
  Encodes `string` as BEP3's weird URL encoded string.

  ## Example

      iex> bep3_encode("a88fda5954e89178c372716a6a78b8180ed4dad3")
      "%A8%8F%DAYT%E8%91x%C3rqjjx%B8%18%0E%D4%DA%D3"

  """
  @spec bep3_encode(binary) :: binary
  def bep3_encode(string) when is_binary(string) do
    string = Base.decode16!(string, case: :lower)
    URI.encode(string, &URI.char_unreserved?/1)
  end

Given an Info hash v1 input of acc3b2e433d7c7475abb5941b5681cb7a1ea26e2, the expected output should be as follows.

%ac%c3%b2%e43%d7%c7GZ%bbYA%b5h%1c%b7%a1%ea%26%e2

However, I haven’t figured out how I preserve the case on the various letters. I assume I need to have exactly the same case-sensitive output as qBittorrent because an ASCII A is not the same as a.

It seems like the way qbittorrent does it, and I apologize for not having the language to communicate this… hex values that can be displayed as ASCII are output as ASCII. Otherwise, the hex representation is displayed.

I wrote this out to make a visual comparison of expected input and output values, because this is the only way I could think of understanding what is happening during the encoding.

 a8   8f   da   59   54   e8   91   78   c3   72   71   6a   6a   78   b8   18   0e   d4   da   d3
%A8  %8F  %DA   Y    T   %E8  %91   x   %C3   r    q    j    j    x   %B8  %18  %0E  %D4  %DA  %D3

On the second line (expected output) See the ASCII Y? That matches up with Hex 54. The three values before that were all not able to be displayed as ASCII, so the hex value was used instead.

Later on, we can see lowercase x, r, q, j, j, x. From what I can tell, Elixir’s built-in URI.encode/1 can’t do this, because like I mentioned earlier, that can only output A-F,0-9. No lowercase!

By the way, https://www.asciitohex.com/ has been very helpful.

Anyway, I am very confused. I’m going to rest now and pick this up in the morning.