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

Re: [802.3_4PPOE] Link section vs "channel" and link segment

On May 15, 2017, at 2:38 PMPDT, Yair Darshan <YDarshan@xxxxxxxxxxxxx> wrote:

Geoff and all,
“A PSE that is in standby and/or searching is not "active".””
Yair: I like the following preliminary definition My current thought is that it should mean a PSE sourcing power and/or a PD sinking power supplied by a PSE.
“We could make it more precise by specifically limiting to an explicitly defined set of state machine states”
Yair: I don’t like to further complicate the state machine. Good text will do the same as in other parts in the spec.


This would not complicate the state machines.
This would not change the state machines.
This would only refer to the state machines.
E.g. A PSE shall be defined as active when and only when it is in either the POWER_ON or the TEST_MODE state.

(Even if we did not define it this way in the body of the standard, we would have to define it in that manner in the management behavior.)


From: Geoff Thompson [mailto:thompson@xxxxxxxx] 
Sent: Monday, May 15, 2017 8:34 PM
To: STDS-802-3-4PPOE@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_4PPOE] Link section vs "channel" and link segment


On May 15, 2017, at 9:04 AMPDT, Heath Stewart <hstewart@xxxxxxxxxx> wrote:
This all has me wondering what active means...
I am all in favor of getting this fully defined dow to the last knat.

Does active mean both PSEs are powered but only one is "active". Note Dual-sig PDs will present a detection signature on both modes.
My current thought is that it should mean a PSE sourcing power and/or a PD sinking power supplied by a PSE.
A PSE that is in standby and/or searching is not "active".
We could make it more precise by specifically limiting to an explicitly defined set of state machine states

Or does active mean powered?
see above.

The word active can be deliberately manipulated to mean whatever a vendor wants it to mean.
Not if we limit the definition in the standard.  If they go outside or beyond what we define then I guess they would be in "proprietary active" land.

(GOT) I was under the impression that 802.3bt allows you to gang one active endspan PSE (Alt A) plus one active midspan PSE (Alt B) simultaneously to get additional power.  
^^^ this is a really bad idea, the IEEE should have nothing to do this type of system
We had lots of discussion about this and frankly I'm not sure where we ended up (Actually, the door is still open on all questions until we establish the scope of Sponsor Ballot).
Whether we sanction it or not, people will do it in the field unless we provide explicit cautions that give them good reason not to (or that is done for us by UL or NEC).
I am happy to join those who would preclude any 802.3 support for such a system.  It would certainly makes things simpler. 
For my interest in the use of the term "link section" it removes the big ugly complicating case.


On Wed, May 10, 2017 at 3:37 PM, Geoff Thompson <thompson@xxxxxxxx> wrote:
I have copied your questions below
(and corrected some spelling and typographical items)
My answers are mixed in
(For those new to the conversation, these comments from Yair (Red) and my responses (Green) are all relative to my presentation from the March meeting,  )
Slide 2
1. I undestand that the term "channel" is ANSI/TIA term for the channel model with 4 connectors as defined in D2.4 page 211.
Why it is not in the list above?
(GOT) Because we write our standards for the international market with an eye towards eventual adoption by ISI/IEC.  That means we use international rather than national or regional standards for reference whenever possible.  In this particular case the definition in the ISO/IEC standard is technically equivalent to the TIA definition so there is no need to cite both.
2. Does your plan is to change also the usage of "channel" and "permanent link" in clause 145.4.9 ?
(GOT) In a word, yes, but it is a little more complicated in my view.  First, the problem with figure 145-38 is that it is a cabling standard diagram and only (costly) correct in that context.  Thus we face two choices (a) to acknowledge it as a stolen cabling drawing (and somewhat incorrect in our context) or (b) to edit it into correctness.  To do that we would have to do the following:  (a) redepict (at least) the end connectors from a single box with one "c" into two boxes (i.e. a mated pair) then (b) move the endpoint demarcation lines to align to the mating interface of the mated pair and (c) change the term "channel" to "link section".
"permanent link" should be changed to "link segment" (not "link section")  

3. Currently Midspans operated per the definitions of 145.4.9. How your 45 comments to change "channel" to "link segment" will affect it?
(GOT) My changes should fix the technical errors introduced by doing that.  We can still use the terms cross connect and interconnect without a problem (but I feel that they are unnecessary).
Slide 3
1. What is the link section model (how many connectors, cordage? cable?)
2. What is the link segment model (how many connectors, cordage? cable?)
(GOT) It doesn't matter.  The tradition in 802.3 (which is an equipment specification group, not a cabling group) has been to specify the media from MDI to MDI (or in the case of PoE from PI to PI) by overall performance of the link segment/section.  How a cabling standard chooses to divide that up inside their standard and between cables and cabling hardware is outside our scope.
3. Why you prefer to use terms (link segment/Link section) from ISO/IEC and not from ANSI/TIA
(GOT) See answer #1 for slide 2 above.
Slide 4
1. I understand that this definition allows Switch without PoE and Midspan to be connected to the same PD correct?
2. I understand that this definition doesn't allow Endspan and Midspan to be connected to the same PD, correct?
3. Regarding (2), what if one of the PSEs (Endspan or Midspans is not active), does it meets the definition above? 
I am asking because we have equipment that work like this for backup purposes.
4. What is your opinion regarding to modify the the definition above to
"Link section: The point-to-point medium connection between two and only two active Power Interfaces (PIs).
(GOT) I think your addition to my definition improves the definition.  It may need more work or a clarifying note to cover the case during the contention process while the PSE PI to be used when the system goes active is being chosen.
 Slide 5
Yair: I like the current one since it is more flexible. It allows me to support my current equipment while the above doesnt allow....which is my main problem unless you modify your fantasy definition to the one that I have suggested in the previous slide. 
(GOT) The current one is not correct.  That is because the "scope" of generic cabling does not include the "application specific connector" (i.e. the RJ-45 that is at each end of the "generic cabling"). 
1.4.248 link: The transmission path between any two interfaces of generic cabling. (From ISO/IEC 11801.) 
As I said above, I like your modification.  We will probably need to discuss this a little more in the meeting.
Slide 9a
1. where from IEEE802.3AF/AT?
(GOT) Again, perhaps we should add the word "active" to the slide.  Otherwise, this is a trick question.  You know the answer much more intimately than I do. 
 Slide 9b
Why it is more complex? This case is covered by IEEE802.3AF/AT and 802.3bt.. (Backoff time etc. It means that there is only one PSE active.
(GOT) I was under the impression that 802.3bt allows you to gang one active endspan PSE (Alt A) plus one active midspan PSE (Alt B) simultaneously to get additional power.  In that case, if we say that there is one and only link section it has (at least) a funny shape.
Slide 10
1. You can't prevent such a case from happening, it can be by mistake and most likely will not work over 4-pairs due to serious unbalance issues (It will fail PSE_Vdiff, Icon-2P_unb, Overload etc)
(GOT) This is not a question.

2. If only one PSE is active, that there is no issues at all. Correct?
(GOT) Correct

3. I prefer using link segment instead of link section. What is your strongest argument against this proposal.
(GOT) Link segment and link section are exactly equivalent in the case of an endspan PSE.
We invented the term "link section" during the development of 802.3af to be able to have a term for the power path when it was not the same length as the data path (i.e. for the midspan case).  We made the decision at the time that the term "link section" was defined to have it apply to either the endspan case or the midspan case.  We COULD have decided at that time to only make the new term apply to the midspan case but that is not the decision we made at that time.  That is not a decision that we can revisit at this point. 
I hope this answers your questions.  My guess is that to go any further on theses issues we need to discuss it during the meeting rather than by correspondence.  I look forward to seeing you in New Orleans.
Best regards,
On May 8, 2017, at 11:22 AMPDT, Yair Darshan <YDarshan@xxxxxxxxxxxxx> wrote:
Hi Geoff,
You have submitted many comments regarding replacing channel with link segment.
I have read your presentation and I have some questions and some concerns.
I would like to start a discussion with you about this much before the next meeting and I am attaching your presentation with questions/comment inline.
Ill appreciate if you can address them.
Many thanks

Darshan Yair

Chief R&D Engineer
Analog Mixed Signal Group
Microsemi Corporation
1 Hanagar St., P.O. Box 7220
Neve Ne'eman Industrial Zone
Hod Hasharon 45421, Israel
Tel:  +972-9-775-5100, EXT 210.
Cell: +972-54-4893019
Fax: +972-9-775-5111

E-mail: <mailto:ydarshan@xxxxxxxxxxxxx>.  

<Thompson_01_0317_with Yair comments.pdf>

Heath Stewart
Design Center Manager, Mixed Signal

Office   (805) 560-7658
Mobile  (805) 895-0499

Linear Technology is now part of Analog Devices.  Learn more.