[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Layer 3.5 is actually layer 4

On 26-jan-04, at 23:27, Masataka Ohta wrote:

UDP being a special case
there will be other special cases for non-TCP protocols including
ones we have not invented yet

UDP is a special case because it doesn't provide reliability or congestion control. There is no reason why UDP should necessarily be a special case for multihoming. The only reason to make it a special case would be because UDP interactions are often very short-lived.

And yes, there will always be special cases, the world is a complex place.

that there is no reason to insist on the fallacy.

Which fallacy?

So, your reduced requirement is merely that: 

	The systems level argument for a layer 3.5 solution is
	that it can cover a TCP case.

which is a TCP specific layer 4 solution.

I don't see any reason why a solution that sits between IP and upper layers couldn't work with all upper layers, first and foremost TCP, but also UDP and the real-time protocols built on top of UDP.

Note that layer 4 protocols can and will share some function
calls to control M6, just as TCP and UDP can share a function
call to compute check sum that it is an issue independent of
layering. We don't have to make transport layer check summing
layer 3.5, only to let TCP and UDP share some function calls.

If you prefer to look at it this way, I have no problem with that. Different people have different outlooks. One man's waves are another man's particles...

Iljitsch