Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: Subject: Survey of lengths of installed optical fiber cable

No one's disagreeing with your point.  I'm just saying that your point is
not encompassing of the potential applications out there....and is
counterproductive.  One of 10GE's objectives is to provide a "standard", "as
cost effective as possible" solution to the masses.  If the masses choose to
use it, it's there business.  Also, if the masses choose to use it, the cost
goes down.  If you discount the importance of 'economies of scale' in total
cost picture, then your naive label is sorely misplaced.

High end proprietary links are going to be idiosyncratic, and many of these
links WILL be proprietary.  But it's very possible they can use the same PMD
as 10GE.  Thereby taking advantage of the cost effectiveness.  Ignoring
these application's applicability to a cost effective solution, in fact
resisting it, is not productive to the technology.....and the concept of
economies of scale.


-----Original Message-----
From: jay.hoge@xxxxxxxxxxxxxx [mailto:jay.hoge@xxxxxxxxxxxxxx]
Sent: Wednesday, May 17, 2000 1:54 AM
To: Edward Chang
Cc: Roy Bynum; Chris Simoneaux; stds-802-3-hssg@xxxxxxxx
Subject: RE: Subject: Survey of lengths of installed optical fiber cable

I disagree.

When designing a proprietary interface, one has the flexibility to
"right-size" the link. That is to say that you can match the data rate,
feature set, etc. to the particular ideosyncracies of the system. I think
that it is unlikely that, in a world of terabit class routers, board to
board interconnects will be limited to 10GB/s. Hence, we are really talking
about sizing in two dimensions. Links can be designed using many of the
same components used in more generic intersystem links. This allows a
designer to optimize price/performance to a particular system. A good
example of this sort of interface might be the purely logic to logic links
implemented using 1x12 VCSEL arrays; no XAUI, no line coding, no frills.

To think that all of the bells and whistles we are talking about adding to
10GBE to insure interoperability, generality of application, etc., come at
a price which can be canceled out by economies of scale when compared to
lean and mean, purpose built links based on many of the same components, is
naieve and unjustified.