Signal vs. Idle debate (was: Here's a new idea)
There are conflicting opinions on where the priorities should be. I'll be glad
to take up the challenge to distract you from your work and make you all wade
through my opinions for a while.
There are a number of customer who have previously purchased Ethernet
hubs/switches/routers and deployed them in their facilities. I believe that
approximately 100% of the installed base fall into this category of investors.
(I suspect that the number of us who are using near death alpha prototype
equipment at home is less than a percent - perhaps I'm wrong.)
Some number of the above sub-group of Ethernet users have "cheaped out" and
either run just two pairs or have decided to alter their plant and run two
Ethernet links over the installed four pair. I would conjecture that the vast
majority of the people/organizations that did either of these thing knew they
were cutting corners. Some potential users may very well have installed other
facilities that they are migrating to Ethernet. These organizations/people may
have installed the facilities in good faith and may now find that they can not
take full advantage of the breath of the Ethernet, they are already precluded
from 1000BASE-T.
Of the power schemes that have been proposed, only the idle pair proposals have
addressed preserving the investment of all of the current 802.3 users by
allowing mid-span power insertion. While the signal pair camp may claim they
are preserving the wiring infrastructure of a proportion of the of the 802.3
users ignored by the idle pair camp, they require every customers - including
those that followed the recommended practices in good faith - to dispose of all
of their existing hubs/switches/routers to make room for the users (or potential
users) who have cheated or sub-standard plants. This is a huge expense to place
on the backs of customers who have properly engineered plants just to allow a
minority the special advantage of not having to come up to spec.
I am of the opinion (and trust me this whole mail is just opinions) that the
real markets for power over Ethernet are not yet discovered, and between now and
then a huge number of Ethernet ports will ship. If we require customers to
replace all their wiring closet equipment, the adoption rate of power will be in
fractions of a percent of the total shipments forever. If we provide a means
for users do not have to fork lift all their capital equipment, we will allow
the power applications to be added incrementally and the market will develop.
The 802.3 has effectively already told the two pair users they are at the end of
the road. We did this with the 1000BASE-T standard. These users should not be
expecting the 802.3 to provide them any new technology on their old plants, we
have demonstrated this to them previously. Why would we turn around and now say
we were just kidding about not supporting two pairs, should we also re-examine
if we can run 1G on two pairs?
I am willing to characterize the performance of powered systems over CAT3 / CAT5
per 568. Are we really saying we want to characterize this as well as every
form of STP/UTP that could "commandeered" for Ethernet as well as every form of
24/26 gauge telephone wire (with and without stubs and / or but slices) ....? I
for one would like a chance at completing a spec.
I would say the idle pair mentality is the most inclusive, it allows use of 100%
of the existing users to upgrade and the signal pairs group is the most
exclusive, allowing no users to upgrade (without a forklift.) Limiting the
scope of study to properly engineered plants allow at least some users to enjoy
the benefits of powered Ethernet while open ended standards will never complete.
The distraction of my opinion storm is over, you may return to your real work.
Mike