Hello Abhi
We need to find a way to move forward. Let us recall the history of this topic
1．
I presented this topic in 19/1988 about two year go, you argued there is TSF sync issue between affiliated APs,
then one link with bitmap was accepted in 21/80, but bitmap leaves space for multiple links
2．
With the help of Kaiying and Minyoung’s work, TSF sync issue was addressed with the proposals of TSF offset
and clock drift requirement. Then you argued it can’t work when I presented the case of multiple links, saying the target wake up time already passed in some links in the call because the starting time of each link is different
3．
After the work of Rubayet, the fig he drew demonstrates that it works well, then you asked how it works, then
Minyoung provided the equation to tell you how it works.
4．
Meanwhile Liwen proposed to reference the tsf of the fixed link to address the framebody change issue when the
packet is retransmitted on the other link, then you argued there is issue when a fixed link is not available (ML reconfiguration)
5．
To address your concern, Rubayet and I worked together and changed it to reference the tsf of the link that indicated in the bitmap (to
which the TWT agreement applies), and kept Minyoung and Liwen’s suggestion. Then you argued it is complex. But
there is
only “addition” in the equation.
Best wishes,
Ming Gan
Hi Abhi
The exiting work already includes bitmap field and can support multiple links, why not use that directly?
To address your concern, I keep “using the TSF of the link to which the TWT agreement applies ”. Meanwhile I also incorporates liwen and Minyoung’s contributive suggestions, Please
refer to the following text
•
If multiple links are indicated in the Link ID Bitmap subfield of the TWT element, then multiple TWT agreements are requested to be setup; A TWT agreement
is requested on behalf of each of the STAs affiliated with the same MLD and that is operating on each of the indicated links.
•
The same TWT parameters are requested for all the links.
•
The target wake time of i-th link indicated in the Link ID Bitmap subfield (TWT_i) is derived from the Target Wake Time field of the TWT element as follows:
TWT_i = TWT_ti + TSF_offset, where
TWT_ti obtained from the the Target Wake Time field of the TWT element is in reference to the TSF time of i-th link indicated in the Link ID Bitmap subfield of the TWT element, TSF_offset = (TSF_0 - TSF_i) and TSF_0 is the TSF time of the setup link
that is associated link ID of the lowest value, where the TSF_i is the TSF time of the i-th link indicated in the Link ID Bitmap subfield of the TWT element.
Best wishes,
Ming Gan
Hi Ming,
Thank you for your response.
>> Before debating the new issue you raised below, I just want to confirm that the answer (Including the fig you requested us to draw) to the issue you
raised in the call is clear and it can work, right?
It’s not a question of whether it will work or not. It is an argument of complexity with no added benefits. Why put unnecessary burden on implementations?
Therefore, the question to ask is: Does the existing solution work under all scenarios? If the answer is yes, then what is the need for an alternative (and more complex) solution?
The alternative has undergoing a few rounds of updates as members have pointed out various issues. Each iteration adding more complexity.
With respect to the latest issue: if we were to go with a ‘reference’ link approach:
- Which link is selected as a reference link?
- Is the reference link same for all non-AP MLDs? Or is an AP implementation expected to track different links as “reference” for different non-AP MLDs?
- Is the reference link required to be part of the ML setup? In other words, is a non-AP MLD expected to keep track of TSF for a link that is not part of its ML setup?
- What if the AP MLD performs ML reconfiguration procedure that removes the reference link? Or is the AP MLD not allowed to remove a link that is marked as ‘reference’?
- Can the reference link be different based on who (non-AP MLD or AP MLD) initiates the TWT setup?
Addressing the above items will add more complexity to implementation and may create a new set of issues.
As I mentioned, the existing mechanism is simple to implement and works for all scenarios. And therefore, there is no need to define another way to do the same thing especially if the alternative adds complexity with
no benefit.
>> For the retransmission, this is not big issue. Anyway, the retransmission for MLO is different from that single link since at least address field should
be changed.
Liwen’s point was about the contents of the frame body. When a frame is retried on to another link, the contents of the frame body remain unchanged (otherwise, you would need to re-encrypt the contents). The TSF reference
(in TWT field) will be carried within the frame body.
Regards,
Abhi
WARNING:
This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.
Hello Abhi
Before debating the new issue you raised below, I just want to confirm that the answer (Including the fig you requested us to draw) to the issue you raised in the call is clear and
it can work, right?
Regarding the new issue, you argument conflicts with your previous one. In previous call, you said TWT setup frame should link specific, then the transmitting link should be known,
however, now you argue “you won't know up-front, which link a TWT setup frame will get transmitted on”.
For the retransmission, this is not big issue. Anyway, the retransmission for MLO is different from that single link since at least address field should be changed.
Liwen’s and Minyoung’s proposals are more constructive, they work for me.
Best wishes
Ming Gan
Hi Ming, Rubayet,
Liwen touched on a crucial aspect of MLO: Due to the nature of multi-link operation, an individually addressed Mgmt frame can be transmitted on any available link. This is one of the key reasons why the standard has defined
a procedure for identifying the intended link for an individually addressed Mgmt frame (i.e., the Link ID bitmap in TWT IE or an explicit IE to be carried in a mgmt frame). In other words you won't know up-front, which link a TWT setup frame will get transmitted
on. Therefore, you can't tie it to any particular link's TSF.
This problem won't occur with the solution already defined in the spec - i.e., the TWT field in the TWT IE is tied to that one link which is signaled via the Link ID bitmap and therefore is agnostic to which link the
setup frame ultimately gets sent out on. If your goal is to setup synchronized TWTs, you can include a TWT IE for each link (with the corresponding Link ID bit set to 1) and the TWT field set to the value based on the TSF of the intended link.
Before adding more "fixes" to your proposal and making more complicated, we need to ask ourselves: Does the existing solution work under all scenarios? If the answer is yes, then what is the need for an alternative (and
more complex) solution?
Regards,
Abhi
WARNING:
This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.
I have a similar comment but the solution is different.
Current 11be TWT can establish the overlapped TWT SPs on multiple links, and TWT SPs are independent of the transmitting link of the TWT Setup frame. It seems that reusing/extending the existing 11be TWT is better instead
of defining another TWT operation.
With the Target TWT Wake Time referring to the TSF time of the link where the TWT Setup frame is transmitted, the frame content of frame needs to be changed if
the retransmission of the frame is in a different link. Is it possible to refer the start time to a fixed link?
Best Regards,
Liwen
Caution: EXT Email
Thanks Minyoung. The following text sounds good to me.
发件人: Minyoung Park [mailto:mpark.ieee@xxxxxxxxx]
发送时间: 2022年4月27日
6:14
收件人: Ganming(Ming Gan) <ming.gan@xxxxxxxxxx>
抄送:
STDS-802-11-TGBE@xxxxxxxxxxxxxxxxx
主题: Re: [STDS-802-11-TGBE] 22/552r2 Individual TWT
If I understand correctly, what you want is to derive the TWT times of the other links from the TWT time indicated in the TWT element transmitted on one of the
links so that the TWT start times across multiple links are aligned. Maybe better to spell that out as follows?
"• If multiple links are indicated in the Link ID Bitmap subfield of the TWT element, then multiple
TWT agreements are requested to be setup; A TWT agreement is requested on behalf of each of the STAs affiliated with the same MLD and that is operating on each of the indicated links.
• The same TWT parameters are requested for all the links.
• The target wake time (TWT_r) of the Target Wake Time field of the TWT element shall be in reference to the TSF time (TSF_r) of
the link the TWT element is sent.
• The target wake time of i-th link indicated in the Link ID Bitmap subfield (TWT_i) is derived from the Target Wake Time field
of the TWT element as follows: TWT_i = TWT_r + TSF_offset, where TSF_offset = (TSF_i - TSF_r) and TSF_i is the TSF time of the i-th link.
(CID#6240 6241)"
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1
To unsubscribe from the STDS-802-11-TGBE list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-11-TGBE&A=1