Closed GoogleCodeExporter closed 9 years ago
I am sorry. I didn't search the docs enough.
This seems implemented by torrent_handle::use_interface()
Please close this.
Original comment by hammered...@gmail.com
on 9 May 2013 at 9:23
I have made some attempts to support this well in libtorrent. the main thing
you cannot do is to bind to a specific device (just IPs). If you come across
some use case that doesn't work in libtorrent, please let me know.
Original comment by arvid.no...@gmail.com
on 10 May 2013 at 2:56
[deleted comment]
I don't have access to multiple connections myself. I will try to expose this
setting in qBittorrent and let end users test it. I will report any problems
here with a new issue.
Original comment by hammered...@gmail.com
on 10 May 2013 at 9:45
I don't think binding by IP is sufficient. If I understand correctly, this
would require fixed IP addresses, and I don't know what behavior would occur if
multiple interfaces each had the same IP address on their respective networks.
DHCP-assigned addresses are probably the norm for most users (especially with
uPNP easing/enabling port forwarding for that scenario). In my case, my IPs on
available networks/interfaces are DHCP-assigned, although I have a reserved IP
on one network and am ever the only client on the other--so my own IPs--when
the networks are up--may be effectively fixed.
* A listing of descriptive interface names would probably be best.
* But one other candidate behavior that might be desirable comes to mind:
Windows monitors network connections and determines when they are unfamiliar.
Windows Firewall users are then prompted whether the network is public or
private, and Windows/Windows Firewall saves their decision for future
connections to that network. This is a level of abstraction beyond interface
to _what is the interface connected to?_. It may be possible to query
Windows/Windows Firewall for that information (but if it's tightly tied to
Windows Firewall, the user may not be running the Firewall/may be running
another firewall...). It should also be possible to use your own criteria for
identifying the current network of an interface. MAC address of DHCP server,
router or gateway might be a most important factor, possibly also combined with
SSID for WiFi networks and IP subnet for all networks. It's debatable whether
you would want to lock that "network identification" to a specific interface or
not, as wireless interfaces are mostly interchangeable, as are wired
interfaces. The "network id" abstraction is more complicated than just using
an interface list, but might be most useful for laptop users that are on
various networks throughout the day (home, work, school, train, cafe, poor
neighbor's wifi). A human-readable approximation of the network ID might be
the SSID and the router name (either advertised by uPNP, DNS, or possibly some
other protocol).
* Actually, a scenario where it would be desirable to include the network
interface as a criteria either alone (as I began my recommendations) or coupled
with the network identification, would be when you are using multiple wireless
or wired connections to the same network. This might come up if the network
has a wideband WAN interface and the user has several independent LAN/WiFi
connections enable to capture as much of the WAN bandwidth as possible, even
though their individual links are each slower than the available WAN throughput.
Original comment by a...@lovetour.info
on 11 May 2013 at 10:20
Original issue reported on code.google.com by
hammered...@gmail.com
on 9 May 2013 at 8:56