Hi Zhen, Thanks for your feedback. I agree there is no such thing as a connection manager in Windows XP. That would be very convenient. However, if you connect to two networks and then explicitly bind a socket to Interface 1 and another socket to Interface 2, then flow aggregation should also work in Windows XP. Maybe the local routing table needs some small modifications? However, the problem about MIF is not only about technical and configuration issues. If there are no user preferences about how and when to use which interface, the OS has a big challenge of making a good decision of how to assign flows to interfaces. If I were Windows XP, I also wouldn't risk binding your real-time video stream to the 3.5G interface if there is also an Ethernet interface available ;-) Dominik On Fri, Sep 4, 2009 at 7:37 AM, Zhen Cao<caozhenpku at gmail.com> wrote: > Hi Dominik, > > I am interested in the issues you raised here. See my comments in line. > > Best regards, > Zhen > > On Thu, Sep 3, 2009 at 8:10 PM, Dominik Kaspar<dokaspar.ietf at gmail.com> wrote: >> Hi, >> >> I have been working on MIF-related topics for a while now, but only >> recently discovered the existence of the MIF working group. Excuse me >> if I ask questions that might have already been sorted out in the >> meeting or on the list. >> >> The MIF charter specifies (among other things) the following restrictions: >> >> - The group shall not develop new protocol or policy mechanisms. > some problems are not clear for multiple interfaces, so the WG would > like to be more informed of the problem before working on protocols > and mechanisms. > >> - No work will be done to enable traffic flows to move from one >> interface to another. > Move of the flows is out of scope, but IMO, source address selection, > which is related to this problem, is under consideration. > >> >> To my knowledge, all major operating systems can cope with multiple >> interfaces and make a decision about using one of them for all new >> traffic flows. I am therefore curious if the group will work on >> assigning new traffic flows to the available network interfaces (based >> on known or measured link characteristics) or if that contradicts with >> the goal of not developing new policy mechanisms? > > But many operation systems and connection managers have not fully > utilized all the available interfaces. For instance, my windows xp can > only send data flows via one default gateway. > >> >> One solution that already seems to work pretty well is Equal-Cost >> Multipath Routing, which (after some non-standardized configuration >> difficulties) can also be set up on MIF end-hosts. At least with >> Linux, a small ECMP routing table entry can have great flow >> aggregation effects when many transport connections are opened. >> >> Best regards, >> Dominik >> _______________________________________________ >> mif mailing list >> mif at ietf.org >> https://www.ietf.org/mailman/listinfo/mif >> >
Note Well: Messages sent to this mailing list are the opinions of the senders and do not imply endorsement by the IETF.