As stated above even if he is doing that its implied, not a direct statement. Assume away. Also, doing this will actually make your connection worse as that connection will always be more latent.
Think of it this way. When you stream a movie from Netflix it is one TCP session. Not 5, 10, etc. So if you load balance to your neighbor your stream will be worse by default. This is not equal cost load balancing, this is a hack that isn't what it seems and there are a lot of people commenting that have, apparently, little to no knowledge about basic network fundamentals.
Try it and prove you've doubled your bandwidth. There are tens of logical reasons why this doesn't work.
Think kd it this way. When you stream a movie from Netflix it is one TCP session. Not 5, 10, etc. So if you load balance to your neighbor your stream will be worse by default.
Like others have pointed out, Linux doesn't load balance a single connection over multiple WANs, so as long as the Netflix stream gets on the best link, it'll be better since it'll have to compete with fewer connections (since some will be routed over the other link).
4 lane highway vs 1 lane dirt road. While combining the two makes the number of paths greater the paths themselves are not equal (my equal cost load balancing remark). So, if the router is not taking link cost into account, which it is not, you'll actually have worse performance over all of your connections because the router will try to balance them equally on session start. This means that 50% of the time you start out going to your neighbors connection and fail or go slow. It's actually more advantageous to wait for a spot on the wide, fast link than it would be to take the alternate path.
If the router was taking into account link reliability and speed (overall cost) it would only give a small percentage of connections to the worse link and only if the main link was saturated from a bandwidth perspective. None of this is happening by round robin load balancing which means more of your connections are worse. This is why routing protocols that have these features exist.
In the general case, sure, though I was replying to your specific example. That said, the software he uses for load balancing (mwan3) does support different weights for each wan connection, it's just a matter of configuration.
Sure. So admit you were wrong first, then explain the defensible position you meant to give.
Your comments are clearly not kind or necessary, and not all of them are even true. You might wish to rethink your commenting strategy, because you're clearly smart and I would rather that your opinion not be lost entirely.