A couple of smaller merges to 'next'

[Jesper Louis Andersen]

Hep!

Some smaller merges to 'next':

jlouis/types - implements some type annotations that makes the
dialyzer far more aggressive

jlouis/duplicate - Preliminary work for checking duplicate torrents.
Not really cooked too well yet and probably contains errors.

jlouis/snubbing - Snubbing support for the rate manager. We will need
this to implement a better choking/unchoking algorithm.

The plan is to work on a globalization of etorrent_t_peer_group. We
need choking/unchoking to be handled at a global level and that is
possible by gloabalizing the peer_group process. When it has been
globalized, I plan on starting implementing the choking algorithm used
in the Mainline 5.2.0 code.

[jesper.lou...@gmail.com]

On Jul 23, 9:40 pm, "Jesper Louis Andersen"
<jesper.louis.ander...@gmail.com> wrote:
> Hep!

> The plan is to work on a globalization of etorrent_t_peer_group. We
> need choking/unchoking to be handled at a global level and that is
> possible by gloabalizing the peer_group process.

This has now been done. Pushed to 'pu' for now.

> When it has been
> globalized, I plan on starting implementing the choking algorithm used
> in the Mainline 5.2.0 code.

This should be fairly easy now :)

[jesper.lou...@gmail.com]

On Jul 23, 9:40 pm, "Jesper Louis Andersen"
<jesper.louis.ander...@gmail.com> wrote:

> The plan is to work on a globalization of etorrent_t_peer_group. We
> need choking/unchoking to be handled at a global level and that is
> possible by gloabalizing the peer_group process. When it has been
> globalized, I plan on starting implementing the choking algorithm used
> in the Mainline 5.2.0 code.

etorrent_t_peer_group was renamed to etorrent_choker. Now there is
only one such process. Next, the bad peer manager is now just a peer
manager and manages starting of peers (for outgoing peer connections).
For incoming peers, it is handled by the acceptors. The change makes
the chokers responsibilities fewer, as it is now only responsible for
choking/unchoking.

The choker uses the Mainline 5.2.0 code at the moment. This looks like
a major step up compared to earlier. We still need some queueing
system to make this really good, however.

A number of small fixes has been stuffed into master as well. These
fix several problems I've been encountering when using the client.
There are probably more, but we'll nail them one at a time :)

Next up: Queueing support, Global bandwidth limitation. We'll also
have to look a bit at the robustness.