My opinion on this dynamic resizing concept is that it is total intellectual bull shit.

How can you decide how far through a "failed" packet that data has begun to fail?

How can you decide that some other size packet is better for "right now" throuh your transmission medium?

How can you decide that the extra protocol needed to "re-negotiate" the transfer process is in and of itself reliable?

What sense does it make to add extra complication that requires data holding buffers on a processor that is already memory constrained?

What do you gain by this scheme? In the end the data either gets to its destination or it does not.

A CRC only indicates, with a finite degree of uncertainty, that the data, handshake or control transfer arrived correctly. There can be NO real way to evaluate subjectively that some other packet size will make it through your transmission medium any better or not!!!

I suggest that if the errors that you experience in your transmission medium are predictable then just make your packet sizes shorter in the first place.

If the errors are unpredicable then the far better way to improve reliability is to redesign the transmission scheme to improve the signal to noise ratio in the channel.

And if the channel you are using is not up the task of reliably getting the data from point A to point B then the fundamentally basic characteristic to change is the data rate in the channel..... i.e. move further down from the limiting bandwidth in the channel.

Michael Karas