RE: 16-bit 625Mbaud XGMII
Jaime,
I don't know if there was a misinterpretation.
As far as the standard is concerned, I strongly believe that there should be
two standardized interface instantiations, XGMII and XAUI. XGMII as the
baseline interface between RS and PCS. XAUI as a XGMII extender. I believe
that the PMA Service Interface and the PMD Service Interface should have no
standardized instantiations, as this will permit some level of flexibility
for implementations and will also allow the market to decide what is the
best solution.
I am not proposing that the standard lists XAUI as an instantiation of the
PMA and PMD Service Interfaces; although, I can see how implementations may
prefer this for various reasons.
Thanks,
Brad
Brad Booth
Intel Network Interface Division
bradley.booth@xxxxxxxxx <mailto:bradley.booth@xxxxxxxxx>
-----Original Message-----
From: Jaime Kardontchik [SMTP:kardontchik.jaime@xxxxxxxxxxx]
Sent: Tuesday, March 21, 2000 10:59 AM
To: stds-802-3-hssg@xxxxxxxx
Subject: Re: 16-bit 625Mbaud XGMII
Booth,
I do not mind specifying a 4-lane additional optional
interface for Copper as long as the electrical specs of
this interface are included as part of the specific
standardized optical PMDs. In other words, there will not be
an universal 4-lane electrical specification over Copper
that all the various optical PMD proposals have to obey to.
Specifically:
1) the 64/66 proponents may define an optional
4-lane Copper interface at 2.578 Gbaud that fits
perfectly with the type of coding they want on the
optical fiber.
2) the 8b/10b proponents may define an optional
4-lane Copper interface at 3.125 Gbaud that fits
perfectly with the type of coding they want on the
optical fiber.
3) the PAM-5 proponents may define an optional
4-lane Copper interface at 1.25 Gbaud that fits
perfectly with the type of coding they want on the
optical fiber.
And so on. Compatibility and interoperability at this
4-lane Copper interface would not be required, since
the coding schemes are different.
Only in this context I would support the additional 4-lane
optional Copper interface. These optical-PMD-specific
Copper interfaces would be very valuable. In particular,
I plan to implement an exposed 4-lane interface using
option # 3 and I consider it very useful to specify
the electrical characteristics of this interface in
the standard, so multiple vendors can use PAM-5 coding
on the fiber and be compatible and interoperable
at the 4-lane PAM-5 Copper interface.
Now, if you just want to transport 10 Gbps data
on a Copper backplane and nothing else (LOM, LAN On
Motherborad) - then you are free to use the coding scheme
you like, including the one proposed by Infiniband. But this
type of application is being dealt with in other forums, not in
the 802.3ae.
Jaime
Jaime E. Kardontchik
Micro Linear
San Jose, CA 95131
email: kardontchik.jaime@xxxxxxxxxxx