[802.3_MAINT] Signs of aging standard, or what to do when flow control fails

[Glen Kramer <glen.kramer@xxxxxxxxxxxx>]

All,

As I said previously, MAC Clients are now in trouble since they don't know when to generate the next .request primitive.

Some of these MAC clients are actually in scope of our standard. These are the flow control, OAM, Multi-point Control protocol (both 1G EPON and 10G EPON). All these clients now can issue frames too often and won't know when frames are lost.

But the more I look at it, the more interesting it gets. Even if the Flow Control somehow knows not to issue too many MA_DATA.requests, it appears that the Flow Control would exhibit now a completely different behavior from what it was before. I'd argue that it is now fundamentally broken. Please, review the attached slides.

Comments?

Glen

Attachment: PAUSE_problems.pdf
Description: PAUSE_problems.pdf