Re: Unified PMD vs. Unified PHY
Please never directly quote things that I never said.
You'll find in my immediately prior communication on this thread that I said:
"The LAN does not transport SONET". This is a very accurate statement. You
somehow misconstrued this and attributed a direct quote to me replacing my word
"transport" with "support".
Of course the LAN supports SONET. Ethernet does so today at 10 Mbps, with Fast
Ethernet and Gigabit Ethernet. Conversion to/from SONET is seamlessly performed
and SONET is NEVER deployed in the LAN.
10 Gigabit Ethernet objectives include direct SUPPORT of Wide Area Networks,
these include, but are not limited to, SONET.
At this point, please allow me to remind you of reflector usage rules that I
personally feel that you are in gross violation of. I will quote Mr. David Law
"Communications are expected to be respectful, dignified, and germane to the
subject of the reflector.
The reflector is not a 'free speech' forum. Subscriptions may be revoked for
inappropriate communications. These include, but are not limited to: recruiting,
advertising, soliciting, spamming, flaming, whining, and disparaging individuals
I strongly object to your personal attacks on me. Our chair, Mr. Jonathan
Thatcher, is directly copied as a warning to you.
Roy Bynum wrote:
> I take exception to your comment below, "the LAN does not support SONET." I
> don't know where you learned the OSI model, so here is a lesson. SONET is a
> layer 1 media transport encoding scheme, of the same function as 8B10B, with
> additional features. Ethernet is a layer 2, data link protocol. Since
> Ethernet is a layer 2 protocol, it should be able to run on any layer 1
> signal encoding seheme that we want it to. Since P802.3ae is full duplex,
> it does not matter what the signal encoding scheme is for LAN applications.
> Since the LAN implimentations are normally directly from data switch to data
> switch, there are no other functional restrictions, on the transport
> encoding scheme such as legacy transport facilities implemenations. The
> PCS/PMA could be anything. It could be somebody's "gobledy goop", as long
> as it does what it is supposed to do. That includes 8B10B only,
> 8B10B/64B66B, frame stuffing/SONET/SDH/frame, 8B10B/64B66B/SONET/SDH, or
> PAM5. In other words, the "LAN" does not care what you put on the full
> duplex point to point fiber link. Making the statement that the "LAN" does
> not support SONET" is either a case of a lack of knowledge, or a severe case
> of hubris. The "LAN", as we normally discuss it, is defined by the 802.3
> Ethernet MAC protocol, not the transport encoding scheme. The "LAN" can
> support anything that we want it to, including SONET/SDH.
> Personally, I do not think that adding ~3% additional overhead to the WAN
> compatible PHY, just so you can continue with an encoding scheme that you
> have an emotional attachment to, is justified. When ~3% average transfer
> rate can be recovered by IPG compression using a frame stuffing scheme, it
> become even less justifed. The real alternative for a unified PHY is the
> WAN compatible PHY, because it will already support the LAN as well, with
> only an average ~2% transfer bandwidth penalty.
Palleeeaaase... Cut the "emotional attachment" baloney, it's getting a little
64B/66B was developed for the LAN PHY to reduce the line rate of the Serial LAN
PHY from an initial 12.5 GBaud proposed by some to a line rate far more
compatible with existing 10 Gbps opto-electronics. Even though this issue is not
settled in 802.3ae, the looming consensus, indicated by the overwhelming support
for the Serial LAN PHY proposal authored by Mr. Vipul Bhatt et. al. is that a
~10 Gbps line rate is desirable over a 12.5 GBaud line rate. The same group of
supporters believe that there is value in a Serial PCS/PMA which provides robust
frame delineation, synchronization, limited run length, in-band control code
transport, direct compatibility with the XGMII and XAUI, does not require frame
stuffing or IPG compression, straightforward mapping to SONET without requiring
payload processing, I can go on and fill your screen, for a mere 3.125% overhead
for a SONET based WAN and NO overhead for the LAN and all other WANs which are
not based on SONET.
The WAN PHY as currently defined is NOT LAN compatible since it requires that a
full SONET to Ethernet bridge be present whenever a legacy Ethernet stream is
> If you are honest about wanting a LAN only PHY at 10.00 Gb, seperate from a
> WAN/LAN compatible PHY at ~9.8Gb, then you can have the LAN with a block
> encoding of 8B10B or 8B10B/64B66B and a WAN/LAN compatible PHY with frame
> stuffed, scramble encoding using the SONET/SDH format. If you are honest
> about wanting a PHY that is compatible with both the LAN and WAN
> implimentations at less than 10% additional cost, then you want the LAN/WAN
> compatible PHY with a modified version of the Nortel frame stuffing using
> scramble encoding with the SONET/SDH format.
I'd rather shoot for a UniPHY with 0% additional cost over the LAN which
SUPPORTS SONET. This is what the survey said. Heck, if the survey participants
want to pay 10% more, I'll gladly take their money :-)
> Thank you,
> Roy Bynum
> > Roy,
> > Thanks for the clarification. It appears that you are now in agreement
> > with Mr. Andreas Bechtolsheim's original statement that the UniPHY
> > proposed by Mr. Howard Frazier, et. al. affects the transport of
> > Ethernet over SONET to the tune of only 3% (3.125% to be exact)
> > overhead. I also agree that this is a compromise with the WAN
> > compatible PHY, which is not LAN compatible.
> > I have a few additional comments below:
> > Jonathan, Question for you below.
> > Best Regards,
> > Rich
> > --
> > Roy Bynum wrote:
> > >
> > > Rich,
> > >
> > > A LAN only PHY that is at 10Gbps transfer rate is exactly what I am
> > > talking about. The proposed "UniPHY" is ~3% less than the original
> > > proposal for a WAN compatible PHY which, as you pointed out, is less
> > > than the original LAN PHY to start with. This appears to be a good
> > > compromise. Full 10 Gbps for the LAN and only 3% penalty for the WAN.
> > >
> > > As far as a unified PHY is concerned, no one has said that the WAN
> > > compatible PHY could not be used in a privately owned fiber systems,
> > > like the LAN PHY.
> > I have absolutely no problem with this.
> > > The WAN compatible PHY just has ~5% less MAC transfer rate. As far as I
> > > I know, there has never been a presentation that would only work in the
> > > WAN, ie. a WAN only PHY. This understanding makes the WAN compatible
> > > PHY functional as a reduced rate LAN PHY.
> > The WAN PHY is not compatible with the LAN since it is speaks SONET. The
> > LAN does not transport SONET. The LAN transports Ethernet (in most cases).
> > Am I missing something?
> > > In other words, the unified PHY that Jonathan Thatcher is wanting is
> > > already the WAN compatible PHY.
> > Jonathan,
> > Can you shed any light on this?
> > > I think that the preferences that came out of the survey that Jonathan
> > > indicates that the majority of the group would prefer the WAN compatible
> > > PHY because it will support both the LAN and WAN implimentations at only
> > > ~5% verhead cost. From what I have been able to determine, the actual
> > > implementation costs seem to be about the same. The second alternative
> > > is seperate LAN only and WAN compatible PHYs.
> > What survey are you referring to? The only survey I know of:
> > http://grouper.ieee.org/groups/802/3/ae/public/mar00/brown_1_0300.pdf,
> > shows a WAN PHY in dead last place and a UniPHY in first place.
> > The second alternative is again a UniPHY with a 10% penalty.
> > The third alternative endorses separate LAN only and WAN compatible PHYs.
> > > Thank you,
> > > Roy Bynum
Richard Taborek Sr. Phone: 408-845-6102
Chief Technology Officer Cell: 408-832-3957
nSerial Corporation Fax: 408-845-6114
2500-5 Augustine Dr. mailto:rtaborek@nSerial.com
Santa Clara, CA 95054 http://www.nSerial.com