IPG variation rule
I would like to propose a precise specificaiton of the IPG variation
concept proposed in:
Haddock_1_0700 gives a table which embodies a specific implementation
which satisfies the concept, but it does not provide a precise way to
test whether other implementations might also satisfy the concept. I
believe that the specification proposed in this Email is consistant with
haddock_1_0700, so this isn't a change in what haddock_1_0700 proposes,
only a different way of stating it. Here is what I would propose:
"The IPG may be less than 12 Idles in length provided that the
accumulated IPG Deficit Idle Count (DIC) does not exceed 3. The DIC is
computed at the end of each Inter Packet Gap as follows:
DIC = max( 0, DIC + 12 - IPG_length)
This rule is necessary to allow SOP alignment of the output of the MAC
in the case that packets are spaced by minimum Inter Packet Gaps.
Without this rule, aligning an unaligned SOP following a minimum length
IPG could only be effected by increasing the IPG length, and this would
change the data rate. Note also that the minimum IPG permitted by this
rule is 9 bytes."
The rule works by ensuring that no more than 3 Idles are stolen from the
IPGs of any contiguous series of packets. Note that the deficit could
last an indefinite amount of time, so the concept of average can't
really applied cleanly to defining IPG variation.