extended messages and ut_pex - too few messages from peers

[overkrik]

Hi, I am implementing very simple client to understand BitTorrent protocol better. Right now my client works like this:

- Open connection to peer(from tracker's peer-list)

- Send handshake with extended bit set

- Read peer's handshake back and send extended message immediately after getting last byte of handshake, I am using following keys in message: m = {ut_pex = 1}, v = "appname", e = 0, p = port

- Read extended handshake from peer and drop connections to peers, which doesn't support ut_pex

- Listen for incoming messages(my client doesn't send anything from this point)

When I start to listen for incoming messages I have around 150 out of 500 peers connected. I can see lot's of "bitfield" and "having" messages coming to me, but there are only 3-4 extended(20) messages in about 10 minutes. Is it ok, or should I get more ut_pex messages? Can this be related to using custom app name, or not using some values of extended message? Maybe peers expect me to send anything?

I am not beside NAT and I can bdecode extended messages just fine, so I think there are no low-level errors in my app, but my high-level logic can be wrong.

[Switeck]

Far as I know, uTorrent only uses Have ALL (I'm a seed!) or Have none (I'm a new peer!) for the fast have extension...which is really a shame.

[overkrik]

Hi, thx for reply.

I was able to fix this problem, in my case it was a simple bug. I was expecting data from socket to be shorter than I want, but didn't expect it to have more than one message.

As for ut_pex, this is utorrent extension(it's ut_ for a reason ) and any modern version of utorrent will send it, helping any client that support ut_pex to get more peers.

[Switeck]

So how many on average are sending/receiving ut_pex now?

(percent-wise, roughly)

[overkrik]

I am not sure that my statistic is good, but in my tests out of 500 peers reported by tracker, around 300 are dead, 50 don't support ut_pex, and 150(most of them are utorrent) can recursively discover around 1500 new unique peers.

[Switeck]

I think you'll find that of the 300 dead ones...probably >50 of them couldn't possibly be real.