So, I’ll jump (back) in here – generally I would object, unless the 802.1 folks can confirm that NOTHING relies on a packet being good. 802.3 networks put a lot of stake in minimizing the propagation of packets with undetected (and likely
undetectable after bridging) errors. The general trend in .1 has been to modify various specs for the unspecified quality of wireless networks. I don’t feel qualified to say that everything would still work if we allow any amount of undetected errors into
the network.
From an architecture point of view, the difference in reliability has proven in the market’s acceptance of 802 networks, and people’s continued use
I have no issue with deleting the first note, (for some applications the performance may need to be better) which is the editor’s suggested response.
The editorial comment is just a reorganization, no issues there.
-george
From: ROBERT GROW <bobgrow@xxxxxxx>
Sent: Tuesday, May 2, 2023 4:40 PM
To: STDS-802-3-DIALOG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_DIALOG] Removing error ratio requirements from Std 802
Perhaps I should have provided better pointers. As George noted, people willing to look at this can look at IEEE Std 802-2014 to read the requirements applicable to 802.3 links. There are two comments on P802-REVc/D1.0 subclause 6.2 (comment
database link can be found at https://1.ieee802.org/maintenance/802-revc/). CID #95 proposes edits (not changing the Error Ratio technical requirements), but CID #123 proposes deleting the subclause.
So I’ll refresh the question to 802.3 participants from the P802-REVc comment resolution group. Would 802.3 participants object to subclause 6.2 being deleted from Std 802.
—Bob
Looks like the 802rev webpage misled me. (https://www.ieee802.org/1/pages/802-rev.html )
It points to a rev as the latest - which turns out is from 2010. In that one it is 5.3... the content is the same though, and is now 6.2 in 802-2014 as you point out.
(somewhere along the way they added new content which changed the numbering)
You're right, the wireless folks don't like this because it makes wireless links look inferior. However, the requirements I see do not apply to them, unless they've changed that in the revision draft. In 802-2014 are only for wired & optical, for wireless
it says basically the media can't be trusted: "For wireless physical media, the error performance within a single access domain is variable over time, and no guarantee of service can be given."
-george
-----Original Message-----
From: ROBERT GROW <bobgrow@xxxxxxx>
Sent: Tuesday, May 2, 2023 11:32 AM
To: George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>
Cc: STDS-802-3-DIALOG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_DIALOG] Removing error ratio requirements from Std 802
George,
I’m slightly confused by you citing subclause 5.3. The topic is subclause 6.2 (same number in both IEEE Std 802-2014 and P802-REVc/D1.)
So my concern becomes would deleting subclause 6.2 leave us without an overarching requirement that is applied to all PHYs. Does the Error Ratios requirements in Std 802 serve us (802.3) in evaluating PHY proposals; or is analysis of detected and undetected
error ratios so entrenched in our culture that we no longer need Std 802 subclause 6.2.
The feeling I got from the wireless participants in today’s call was that they are happy not having Error Ratio specifications in Std 802, and with the complexity of their encodings and one wireless type only having a CRC-16 (and thus much worse false packet
acceptance than that specified for wired/optical links. During the discussion it was also noted that the undetected error ratio (item b) is unmeasurable, only mathematically calculated assuming the link meets item a requirements.
All participating in the P802-REVc discussion today on this topic, I think, understood how error ratio quality has changed since 1000BASE-X 8b/10b PCS error multiplication (some link errors producing invalid symbols some producing multiple data bit errors).
Similarly there was recognition that, FEC or other error correction capabilities might change the effective link error ratio for the data that is then CRC checked by the MAC, and that the current text does not cover error correction in the current error ratio
text.
Do I read between the lines correctly, that you would not object to subclause 6.2 being deleted for the next P802-REVc recirculation ballot because we have been specifying the appropriate requirements in recent PHY clauses?
—Bob
On May 2, 2023, at 10:14 AM, George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx> wrote:
Bob - it's not just the tradition of 'life of the universe' for false packet acceptance, but 802.3 is full of frame loss (or in earlier test BER) requirements. Most (if not all) of the recent phys have some kind of a frame-loss-rate specification in their
clause.
It is stated in different ways in different places. Not sure whether this was your concern, as the "Error ratios" in 5.3 of 802rev is stated somewhat ambiguously both interms of the probablility that "a transmitted MAC frame is not reported correctly" (item
a in 5.3), as well as the probability that an MSDU delivered contains an undetected error" (item b in 5.3). While "a" would count frames that are marked errors, "b" would exclude (be tolerant of) them. Which are we doing? The big issue here is how FEC is
considered.
-george
See, for example:
Clause 92:
Differential signals received at the MDI from a transmitter that meets the requirements of 92.8.3 and have passed through the cable assembly specified in 92.10 are received with a BER less than 10–5.
For a complete Physical Layer, this specification is considered to be
satisfied by a frame loss ratio (see
1.4.344) less than 6.2 × 10–10 for 64-octet frames with minimum interpacket gap.
Others are more explicit on the requirement:
Clause 95:
95.1.1 Bit error ratio
The bit error ratio (BER) shall be less than 5 × 10–5 provided that the error statistics are sufficiently random that this results in a frame loss ratio (see 1.4.344) of less than 6.2 × 10–10 for 64-octet frames with minimum interpacket gap when processed according
to Clause 91.
If the error statistics are not sufficiently random to meet this
requirement, then the BER shall be less than that required to give a frame loss ratio of less than 6.2 × 10–10 for 64-octet frames with minimum interpacket gap when processed according to Clause 91.
Clause 97:
Differential signals received at the MDI that were transmitted from a
remote transmitter within the specifications of 97.5.3 and have passed through a link segment type A (specified in 97.6.1 and 97.6.3) are received with a BER less than 10–10 and sent to the PCS after link reset completion. This BER specification shall be satisfied
by a frame loss ratio less than 10–7 for 125-octet frames. Operation on link segment type B is optional. If supported, the frame loss ratio shall also be met for link segments specified at 97.6.2 and 97.6.4.
Similarly, there are alien crosstalk tests which are specified in terms of meeting a frame loss ratio...
-----Original Message-----
From: ROBERT GROW <bobgrow@xxxxxxx>
Sent: Tuesday, May 2, 2023 9:40 AM
To: STDS-802-3-DIALOG@xxxxxxxxxxxxxxxxx
Subject: [802.3_DIALOG] Removing error ratio requirements from Std 802
Colleagues:
Today, working on P802-REVc comment resolution, a proposal was make to delete the error ratio text from Std 802. My recollection is that we don’t have these requirements in Std 802.3 but have used the requirement in Std 802 as the basis for analyzing if PHY
proposals are acceptable for basic PHY link error rate and for undetected error rate at the MAC service interface.
The current Error Ratios text for wireless networks does not include normative requirements.
I’m asking those more expert than I on this for reaction to this proposal to delete subclasses 6.2 Error Ratios from P802-REVc/D1.0.
—Bob
______________________________________________________________________
__ To unsubscribe from the STDS-802-3-DIALOG list, click the following
link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-DIALOG&A=1
To unsubscribe from the STDS-802-3-DIALOG list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-DIALOG&A=1
To unsubscribe from the STDS-802-3-DIALOG list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-DIALOG&A=1
|