Jonathan,
I would rather not base this discussion around the product offerings of
one (albeit large) vendor. I don't think it is an appropriate topic. If
anyone is interested in particular product lines from a particular
vendor, please feel free to contact me offline :-)
The reason that the discussion got onto cut-through switching is that
it already provides a means to minimize latency within the current
standard. I suggest that cut-through switching used to be popular a
long time ago (the switch that you mentioned was launched in 1994 &
the last model released in 1998). I believe that the current market
demand for cut-through is zero. It is possible that latency will become
so important that cut-through will see a resurgence. Let's reopen this
discussion in 2 years; if cut-through has gained a significant market
share in that time frame then I will concede that there may be a demand
for even more aggressive measures to reduce latency. I won't hold my
breath.
OAM-in-preamble did not propose a change to the PCS coding and was
rejected by 802.3ah - I cannot see how you can use that proposal as
support for changing the PCS coding.
Hugh.
Jonathan Thatcher wrote:
Hugh,
Interesting. Should I assume then from your note that
Cisco is end-of-life-ing all switching methods other than
store-and-forward? Funny, I thought 802.3 standards work for done for
future markets, not current markets.
Regarding OAM-in-preamble, you are right, it was
rejected as the method to bring OAM to the masses. But, had there not
been another "more preferred method," it would have worked just fine,
right? The fact is, 802.3 would have made sure it did. If 802.3 did
preemption, it would work just fine with legacy gear also.
jonathan
Jonathan,
It's odd that you should choose an End Of Life announcement (issued
about 2 years ago IIRC) to support your argument that cut-through
switching is all the rage. Please show what proportion of the *current*
market is demanding cut-through. It was all the rage at about the same
time that ATM-to-the-desktop was the future of networking.
Also, three points about OAM-in-preamble: it did not require new
codepoints; existing and compliant PCS implementations would have
ignored it (it's the preamble); it got rejected - does that tell you
anything?
Hugh.
Jonathan Thatcher wrote:
So, the following is wrong?
(many more similar articles can be found
searching Google)
If EFM had adopted preamble-based OAM,
would this have required new PCS codes and a significant number of "why doesn't it work" support calls? If
so, why was it proposed?
Again, this isn't that hard.
|