802.3z minutes September 8-11, 1997 London, UK. Morning Session September 8. Introductions Theatre Information 802.3z Working Group ballot results (See Howard's Slide Set) Draft 3.1 Working Group Ballot Opened 24 July, Closed 2 Sept Results Voting Members Observers and Liasons included 193 Voting Members 161 Returned (83.42%) 190 Returned 112 Approve (80.58%) 131 Approve (80.86%) 27 Disapprove 31 Disapprove 22 Abstain 28 Abstain Dr. Howard Johnson gave presentation on comments. (See Howie's Slide set) Covered comment submission and resolution process and comment statistics. Comment statistics: 1283 Total 201 Technical Required 241 Technical 841 Editorial Then distributed print outs of comment database. Feedback/Concerns with Electronic Comment Process Very efficient for processing comment submissions. Web most automatic, most emails needed some editing for syntax errors, eg. missing colons. Scott Mason: Would like to confirm content of comment before it gets into the database. Example old text hidden behind scroll bars not seen and then gets submitted. Perhaps cgi script could show what it is about to mail before it does so. Howard Frazier: Minimum change is that Howard has requested that submitter will be cc'd on email to database so submitter will quickly get to see what is going into database. Some comments being duplicated in database (two copies of same comment from same person in database) Perhaps email'd twice or duplicated in Howie's processing. Timeline(Slide): To continue to meet March 98 Std complete goal, need to resolve technical issues and do final technical changes this month. We have budgeted for 2 recirculation ballots, if we go beyond two rest of schedule is difficult to achieve. Comment Resolution is our purpose for being at this meeting. (Slide) Advice: Do a quick triage, identify key comments that will take discussion and work. Then start work on easier technical comments to make progress on resolving list. Keep track of major issues for Thursday or those that cross groups. Big Ticket Items, Cross group issues: To be discussed in breakouts Affected Clauses: 4 Owner: Frazier Comment ID(s): 531 Description: Behavior in response to late collision at 1000Mb/s. Plan: Discuss options in breakout. Affected Clauses: 1 Owner: Grow Comment ID(s): 707 Description: Should GMII be "highly" recommended. Plan: Discuss in GMII. Affected Clauses: 22,35 Owner: Grow Comment ID(s): 879 Description: overlap of clauses 22 and clause 35. Plan: Discuss in 802.3z. Resolved in 802.3z by motion Y: 50 N: 0 A: 9. Discussion summary follows. Currently 37 is written independently except uses register set of 22 and modifies register set in 22 as needed. Discussion was primarily on need for two separate clauses due to difference in electrical interface and handshake method. Annex to 22 instead of separate clause would result in one clause that has two signalling methods would not be preferrable. Method of reusing register set and management interface leverages areas that can be common. We could clarify the role of the two clauses and review text to reduce the overlap of the two. Motion 1 : Keep current overall structure of clauses 22 and 35. Perform changes as necessary to reduce overlap of clauses 22 and 35. Moved: Dineen Second: Albrecht Technical Motion Y 50 N 0 A 9 Motion Passed. Affected Clauses: 35,36 Owner: Grow Comment ID(s). ? Description: GMII timing assumptions vs. Requirements. In PCS timing contraints are described as assumptions, should they be specified as mandatory? Plan: Discuss in GMII. Affected Clauses: 30 Owner: Kaul ? Comment ID(s): 1089, 1067 Description: Repeater function Definitions Plan: Discuss in breakout. Affected Clauses: 30 Owner: Kaul ? Comment ID(s): 639 Description: Expand functionality of "Pause" management attributes. Plan: Discuss in breakout Affected Clauses: 31 Owner: Frazier Comment ID(s): 553 Description: Pause timing. Plan: Discuss in breakout Affected Clauses: all Owner: Johnson Comment ID(s): 666 Description: Not use the phrase "Gigabit Ethernet" in the standard. Plan: Discuss in breakout Then Rich Taborek, presented the PCS big ticket items. The majority are actually items that need to be discussed in the PCS group, but did not qualify as 802.3z big tickets since they did not cross clauses or have major impact. PCS Big Ticket Items: Next page protocol, ack2 protocol. Do we get rid of next page protocol ? 1) Redraw state machines. 2) Operation when one link is set up for auto negotiation and the other for manual 3) Packet processing during errors. 4) Deletion of RD and transition columns 5) Link status mapping to link status or link ok state 6) PICS changes 7) Signal Detect 8) RPAT pattern changes 9) Too quick entry to AN upon receive error 10) Ordered_set/code_group definitions, affects mostly epd 11) Exposed/abstract interface wordings 12) State diagram conventions diff in clause 38 and ? 13) Running disparity text adequacy or diagrams to be added. 14) Comma detect, enable comma detect, code_group alignment text 15) Asymmetric pause effects on other stds and projects. Is it transparent to 802.3x text that says Asymmetric pause is not allowed. 16) Pause AN resolution. Item 1 was declared a bit ticket item. Item 8, signal detect crosses clauses and will be covered Tuesday morning in 802.3z. Affected Clauses: 37 Owner: Taborek Comment ID(s): 976 Description: To NP or not to NP. Plan: Discuss in breakout Short discussion about how to review technical changes then occurred. Scott Mason: Question about procedure for review of technical changes. Howard Frazier: New draft will be available beginning of October which will include all changes from working group ballot resolution. Technical changes and any unresolved technical required comments will go out for a recirculation ballot with the new draft. If additional technical changes are needed as a result of recirculation or additional votes change to disapprove we will need a second recirculation ballot. Jonathan Thatcher. 8 significant items, some were declared big tickets. 1) New Chair and editor for group. Was deferred to later in meeting. 2) Affected Clauses: 38 Owner: Thatcher Comment ID(s): 754 Description: Modal Bandwidth Issues, Differential Modal Delay (DMD) Plan: Review new info. 3) Affected Clauses: 38 Owner: Thatcher Comment ID(s): 417,161, 159, 1219, etc. Description: Link Budget Plan: review proposal for restructuring link budget, mirror 15.3. 4) Affected Clauses: 38,39,36 Owner: Thatcher Comment: Lots Item: Signal Detect. Action: Review this afternoon in PMD subgroup and then have a joint z session at 8:30a Tuesday. 5) Affected Clauses: 38 Owner: Thatcher, Comment ID(s): 163, 606, etc Item: Annex A,B,C,D to keep or toss Action: Discuss in breakout, have motion to set objectives for annexes. 6) Affected Clause: 38 Owner: Thatcher Comment: 951, etc Item: Short wavelength is specified for 770-860. In some places 850 is used as a generic term. Discuss in breakout Affected Clause: 39 Owner: Thatcher Comment: 611, 204 Description: Electrical Attenuation vs. Freq. Plan: Review Proposal from Ed Grivna. Affected Clause: 39 Owner: Thatcher Comment: 200 Description: Remove DB-9 Type 1 Connector. Plan. Discuss. There was discussion on this topic at this point. The commenter felt that if we had a single CX connector rather than the two that are currently supported it would simplify life long term due to a reduction in the number of patch cords and product variations that customers and manufacturers would need to deal with and be be expected to support. This was also related to comment 1227 which was that the current DB9 specification is not sufficient for the CX electrical requirements. Question from the floor. What is the availability of HSSDC product? AMP and Berg connectors and cables are available. Motion: That the Style 1 (DB-9) connector be removed from 802.3z. M: L. Miller S: P. Kelley Discussion. Scott Mason : Believes we should preserve consensus. Encourage vote against because we haven't seen new info that overturns the decision of the previous meeting. Tom Dineen: Change of this size could result in collateral errrors in the document as edits occur. Howard F.: Why do we have two connectors? November voted in DB-9 and HSSDC had not started standardization. In March 97 had discussion of new connectors. Saw info that HSSDC standardization was in progress. But we did not decide to replace the DB-9 connector. Pat: If we vote to remove the connector there is a risk of loss of consensus, we should include connector people in 802.3ab before we remove supported connector. Howie J.: Concern was he wanted a standardized connector. Method by which we conducted connector discussions, gave lots of advanced notice and connector vendors were heavily involved. They thought we were making a final decision. Howard F.: Decisions are not final until standard is approved. Paul Kelley: Simplifies ordering for customer by reducing permutations of products. L. Miller: DB-9 connector electrical performance above 3MHz not spec'd Ed Grivna: Connectors are manufactured in a lot of different grades, electrical performance is available at several levels. Jonathan Thatcher: This part is shipping and is working. Discussion is not based on technical viability of this connector. The rational is market confusion of multiple connectors. DB-9's that meet higher electrical performance are available by reference to other specs. Scott Mason: Use physical info for IEC, electrical specs from 802.3z. so 3MHz spec is not an issue. Motion taken as Motion 2: Motion 2: That the style 1 db-9 conn be removed from 802.3z Moved: L. Miller Second: P. Kelley Technical (75% to pass) Y 16, N 15 A: 23 Motion Fails Howard is concerned that the high level of abstains is an issue. Group did not really make a decision. Motion 3: Keep the Style 1 (DB-9) and Style 2 (HSSDC) connectors in clause 39. M: David Cunningham S: Mangin Howie: Is removing the connector an optimization late in the process? If we make a change, it could trigger a no vote on recirculate. Is this enough of an improvement to risk that? Technical (75% to pass) Y: 42 N: 8 A: 9 Passes. Steve Haddock: reviewed the Repeater/MAC issues. Affected Clause: 41 Owner: Haddock Comment: 587 Description: Add annex 41A to explain derivation of values Plan. Discuss in breakout, propose resolution to 802.3z on Thursday. Typical process has not been to show derivation of numbers but it is a technical required comment End of big tickets, now on to other business. Jonathan Thatcher has tendered resignation as chair/editor of clause 38/39. During course of meeting, PMD task force should come up with nomination for replacement. Task force could have a separate chairs for 38 and 39. Also need editors which don't have to be chairs. Thursday wrap up meeting to affirm or elect sub task force chairs. 1:30 resume in task forces. Tuesday morning meeting resumed to discuss signal detect Jonathan: relevant comments. Comments related to making signal detect required and comments related to specifying transmit disable power spec. 1178, 491, 980 on p. 83 618, 756 on p. 86 326, 327 on p. 87 Summary of change: Make signal detect mandatory for fiber and copper and specify a transmit power for a disabled transmitter that is low enough for signal detect to be off for a disabled transmitter at other end of link. Signal detect does not guarantee known good spec'd BER operation simply that the line is receiving signal. Proposed criteria. If received signal power is below -30dbm, the signal detect is False. If power is above -30dBm and receiving a "good" 8B/10B stream the signal detect is OK. If power is above -30dBm but receive stream is not "good" 8B/10B code, the signal detect can be either OK or false. The reference to a "good" 8B/10B stream was fairly confusing at first for the group. Basically when the receivers are getting a stream with reasonable transition density and running disparity as would be expected for a valid 8b/10b stream, then signal detect should be true. There is no need for a 8b/10b decoder/tester just that the assumption of detecting good signal depends on the stream seeming reasonable. If the stream is not reasonable the parts don't have to detect that and go to a particular value. To complete this spec we also need to spec the launch power of an off transmitter and added activation of a transmit disable, so that an off transmitter will generate signal detect false at the other end. The response times to detect signal detect changes are not specified. Discussion of how to update Jonathan's table to make it more understandable. Jonathan asked: Is there anybody that is dissatisfied with this and will comment against this? No one indicated they would comment against the change. Question: Can a device sending out C codes to a powered down device get a signal detect due to hearing its own transmit with crosstalk. No. Eliminating false positives and false negatives like this are a system level problem. The product designers should ensure that their designs control crosstalk, power supply noise etc., of transceiver, cable, etc to eliminate invalid indications. Is response time requirements not specified an issue? There was some concern about it might be too long. Other standards efforts have found it very difficult to spec response time on something like this. There was a request to spec it so that one could give driver writers a delay. Concern about details need and a hysteresis spec would also be needed. How about a second? Typically parts are much faster, most parts provide ample response time so we may not need a spec. Howard F.: Pick a part that has desired response time, let customer decisions filter out parts that have too slow of a response time. Ed: This is not an interoperability issue so we don't need a spec here. Consensus was to leave response time unspec'd. CX signal detect. ok : min diff sensitiivity <= received signal, 8b10b <= max diff sensitivity fail <200mV p-p off tx power less than 70mV 200-400mV for sensitivity. Some concern about being underrepresented by copper folks at this meeting. At first glance this is ok, by Colin Whitby-Strevens of SGS Thompson Define new spec 70mv off. Delete not implemented Del Hanson: Not implemented in current production serdes chip. Will need to be done outside until new versions of serdes that include drivers are available. SGS: Business issue: Need to roll design and pin needs to be selected. To do sig detect. Not a technical problem at 25m no margin left in a passive system, serdes to serdes with a cable between at 25m jonathan thought we need a driver ed g. thinks we don't need one. Jonathan: If encroaching on worst case on both ends, no margin left to indicate sig detect unless you need amplifier, ed: not an issue david c: need on chip equalizer or something else Summary of above lines, there is a concern that if you build a separate external detect circuit is there enough signal to drive this circuit and the serdes receiver or does one need to do a different design on the board. Motion 4. That the signal_detect function in clauses 38 and 39 be changed from optional to mandatory following the definitions developed by the PMD sub task group. M: Ed Grivna S: Jim Tatum Linda: question: is there a spec for polarity and voltage thresholds: Howard: not spec'd, it is a logical interface. Up to implementer to interface to this. Will vary with part, example 1x9 and GBIC's will have interface Mark S. Concern about low cost optical transceivers: ?: Not an issue. Colin: Would like meeting to strongly enforce desire for signal detect in 39 by having it be a separate motion. Motion by Colin to delete "and 39" died without a second. Back to original motion to add signal detect. Tech motion >75% Y 50, N 0 , A 1 Motion passes Howie J:if one end is in loopback, other end could see valid signal wants clause 37 group to consider whether that causes difficulties. Howard: If PMD group needs support from PCS group ask them at this meeting PCS group pointer to PMD since signal that would affect their code. Break at 9:40 resume at 10:00 in task forces. Resume Thurs 1:35pm Sub task force officer proposal for PMD groups. Clause 39 (1000 BASE-CX) Chair: Ed Grivna Clause 39 Editor: Ed Grivna Clause 39 Assistant Editor: Roger Ady Clause 38(Optical PMD SubTask Force) Chair: Del Hanson Clause 38 Editor: Ed Grivna Clause 38 Assistant Editor: Roger Ady Approved by acclamation Howard and group thanked and acknowledged Jonathan Thatcher for his significant efforts and work as PMD editor and task force chair. Then we had a very short presentaion from Ken Taylor Boston Optical Fiber. I don't know if slides are available on web site. His email if you need further info is ktaylor@bostonoptical.com. Plastic Optical Fiber (POF) was not covered in this draft but he expects to propose it after March. He has a group in 802.8 working on POF use in 802 stds. (to establish suitability of POF for application in 802 networks.) Have an interim meeting in Atlanta regarding this topic. Task force summaries. Howie chaired 1,2,34 123 comments. Covereed all TR and T comms dealt with all but 1. Editorials distributed. If you have a concern about any of his editorial responses let him know.. The rest will be his discretion. Big tickets: Comment 521: Shimon Muller comment to highly recommend GMII to allow maximum flexibility in intermixing PHY's and DTE's at gigabit speeds. No objections to going along with it Comment 666: Paul Nikolich saying that gigabit ethernet is industry trade nomenclature, should use 1000 Base X throughout doc. Proposed response In comment database. geoff: we will use the term 1000 Base X when appropriate, however the group will continue to to use the term "Gigabit Ethernet" as the umbrella term. No other objections to this response. Comment 553: Support for full duplex flow control at 1000 Mb/s has implications on the timing considerations Presently when device gets a pause frame, it gets to finish sending what it has, with one pause quanta (512 bits) allowed for reaction time. Delay through PMD is almost a pause quanta leaving little additional time for logic. Task force that looked at system timing for repeater and mac oriented activities agreed that timing was tight and we should acccept the change to allow two pause_quanta. no other objections Comment: 1223 Geoff Thompson: effective modal bandwidth is undefined. Resolution: Annex 38A usage of effective modal bandwidth has been deleted Any objections? geoff: will not sign off yet and would like to keep it open until Santa Clara no other objections, go ahead with this resolution, but not sign off yet Thanks to the editors, thanks to the database editors. Get resolutions to Howie by midnigh Friday. Howie will mail out TR responses to voters by this weekend Will post comments and resolutions to web. Before next meeting will print out paper sets Comment sort order discussion. By clause and line number or id number by page and line number lots of votes by id one vote So do comment resolutions by page rather than by comment. Assign a new number that is a sequence by sort that it is published in . index sheet with seq number, considered too much work. best with page and line number to work on same area at once scott: if page and line number , look at sort and make sure it really is sorted that way. howie: . used clause , topic, subclause (thought subclause more reliable because pg and line number was more freq in error) but result was some comments were out of order. MAC clause: clause 4 Comment: 531, Shimon Muller, Rich Siefert: Commenters were concerned that a late collision is mapped to excess collision error. This is misleading since excess collision is not what happened. Result, change name. Change excessiveCollisionError to TransmitError (excess coll or late collision) summary:Closed all TR's have some signatures , all T's and all E's that Howard felt were significant. Will post editorial responses to web. Clause 3-5 all resolved. Samesh Clause 30 120 comments, 10-15 duplicate, ~40 T or TR, 2 big tickets more discussion, closed all but 2 T's and TR's, the 2 unclosed were the big tickets. Wants to make a proposal on reflector and get input first Big tickets: Comment 1143 repeater function definitions Commenter Scott Mason : carrier and carrier event resolution Scott thought the symbol error during carrier counter was getting corrupted because the /V/ generated during a collision would increment it as well as a noise generated symbol error. His hope was that looking at the symbol error count and the fcs error count and if there was a difference in the two it could indicate a problem elsewhere in the system. Actually this won't work because noise can convert a valid symbol to another valid symbol but still generate a fcs error. David Law will post a proposed resolution to this issue to the reflector. Commenter: Pat Thaler 726 Comment for range of auto negotiation abilities relative to pause. Discussion on this comment was related to the need to support devices that request asymmetric pause operation and are not able to operate unless it was received. The group did not come to a consensus on resolution. Editors, task force chair and commenter are to work offline to resolve this. Howard: comment 518. Late collisions that occur in inter frame spacing are not reported to mgmt. Group didn't feel it was necessary to change this. Closed 201 of 221 clause 6 all closed clause 22 - 1 TR technical (big ticket 50 0 0 ) still open because need to write explanatory text regarding 22,35. Will add text to better clarify, the two. 1 Technical needs text to be written regarding usage of registers and ANComplete. clause 35: 7 - E, 1 T, 10 TR will point out 7 editorials worth showing to group/relector, 1 tech that needs text, 1 tr needs text, 9 tr's waiting for sign off but responses completed. Any concern they won't be signed: only concern big ticket, rest believes commenter can be convinced Open TR's 710 - needs text for tx and rx to be written 879- response: by motion. Motion passed on 9/8/97 Keep current overall structure of clause 22 and 35 perform changes as necessray to reduce overlap of clause 22 and 35. Moved: Dineed Second: Albrecht Y 50 , N: 0 A: 1 >= 75% pass working on text to clarify relationship of the two in both clauses. 707 : accepted: made GMII timing assumptions in clause 35, to be mac timing constraints Will post to reflector list of editorial comments that were closed. clause 36, 37 : rich taborek 375 total 62 tr , 50 accept 4 reject 4 withdrawn 4 in process 56 tech 257 editorial 65 19 PCS big ticket (See earlier in document for issues being addressed.) next page. Group decided by straw poll 20 y 3 n to retain next page. Bit ack2 acknowledges messages from the link partner's previous next page exchange. Will redraw state machines: Direction of changes accepted. Aliases to shorten long names and long multiple term equations to short name. Straw poll 22 to 1 to put in informative annex with example for running disparity method Jonathan Thatcher: big tickets resolved link budget hit 38.10 and 38.11 added a text/diag similar to 15.3 to better define a link and channel bulkhead to bulkhead signal detect: resolved Tuesday . did not undo Tuesday result, database has changes. modal bw issues, dmd. Left open, defined and set up ad hoc group between now and November meeting to accomplish tasks. David Cunningham to lead. Gathering data for November meeting. What is going on with DMD issue? Jonothan gave an example. Send light down a fiber. Some photons spiral around edge of fiber and avoid travelling through center of pipe. Others bounce across full diameter of fiber as exptected. Those that go through center of pipe get there faster than those that spiral and stay on side. Difference in delay depending on path, rogue fibers that result in delay and get two sigs separated too far. Howard: Will we have a detailed plan for testing/analysis/solution characterizations for Santa Clara? Yes. DMD - differential modal delay Geoff: where does it apply. applies to multimode fiber short and long wavelength and 50u and 62.5 u fiber. Howie: need to characterize what is out there, what could we do to help collect info on what is out there? Jonathan: If every company could find three fibers from where they are installed of at least 300m , . David and Jonathan to discuss how to send us your old fiber. Steve J.: What is it applicable to ? Thinks there is not evidence of what it is applicable of longwave over 62.5. Jonathan does not agree. IBM jitter presentation may be smoking gun that shows we have seen this before. Does this affect fiber from 1 mfg or more than 1 mfg. Yes, expect all fibers could see this effect. Annex a,b,c,d. removing annexes a and b (a: mathematical model used for parameter tradeoffs, b : link lengths and modal bw.) Bob G.:Why remove b. Answer: Contents of b are very dependent on results of dmd work. Once get data from dmd investigation will determine whether it makes sense to put annex back in. 780 vs. 850 doc will resolve text to address issues that full range is available. Electrical Attenuation vs. frequency. Transmit and receive eye at bulkhead for copper assembly . distances and equalization. (?) Remove db-9 : we kept it. TR commenter does not agree to rejection of comment PMD subtask: Del to resolve comm 754 on modal bw 2) bob friedman to determine status of tdr measurement std and provide ref 3) ed grivna to provide intl std 4) 268,258,270,269 references in clause that need to be in a ? All other comments addressed Use of Bessel Thompson filter for measuring eye: what did you do? Called out in 2 places in 2 clauses. Test eye mask and rise fall and jitter, filter was referenced, for those from fiber channel transmit eye mask and jitter measurements have left it as was. For rise/fall loosened wording on filter requirement to can use different or no filter to meet spec. Steve haddock: 1 big ticket. Repeater section a breeze , resolved t, tr, eds Shimon asked for informative annex to explain derivation. Real concern was how you derived timer values, added text to timer definition to describe.. No repeater sub task force meeting for Santa Clara Disposition of comments: delay motion on comments until next meeting. Wait until fully resolved before taking a motion. Geoff: announce that there will be an 802.3ab meeting in Irvine about a month. Check email for details. 802.3ac will meet the day after our Santa Clara meeting. Plan for next meeting: Schedule: 9/30,10/1 interim plan: finish comment resolution, finish tr's, prepare recirculation ballot. Lunch is provided at meeting. Recirculate ballot targeted for 10/14 to 11/5. TR's not approved have to be recirculated, comment and response go out Expect 300-400 comments to work on from recirc ballot. Plenary 11/10-14. Recirculation ballot rules: Default is that vote is same as working group. If voted to approve at wg, no need to reaffirm approve. Working to convert disapprove voters to approve expect to ask for sponsor pending approval of second recirculation. .3ab: report: Howard F.: they met Monday and Tuesday. Working on analysis of baseband and pam 5x5 (level 1 proposal) as basis yes on both. Now need to start looking at forward error correction schemes, realized not at consensus so go off to work Plan to pick forward error correction scheme at October interim. Distributed their blue book, avail on 1000t web site 10/13-14 in Irvine for interim mtg. Howard will forward to hssg reflector. Group passed an objective that they would make a agreement on 4d. Chair's note, added to minutes after the fact: The chair was remiss in not recognizing and expressing thanks to 3Com Corporation for the exceptional quality of the meeting facilities. Lisa Salo and David Law deserve special thanks for their hospitality and attention to detail. In all respects, it was a most pleasant and productive meeting. Adjourn.