Remy, Qin, all, I'd like to make sure we don't overdesign here.
The RELOCATE command, as defined now, considers that all cells to be relocated are equivalent: they have the CellOptions, and are part of the same bundle. If the first cell cannot be relocated, the second cannot either. What is in the text is perfectly suitable for that use case. I'm not at all convinced about the use case you are discussing here. If there are 2 logical groups of cells, and different restrictions on the candidate cells for each, then you just issue 2 independent RELOCATE commands. Let's not cram additional features into this, it's already complex enough. If there is a very compelling counter-example, in a very specific application case, then it's completely possible to redefine the format of the RELOCATION messages in the SF, as specified in the text. Thomas On Tue, Sep 12, 2017 at 3:30 AM, Liubing (Remy) <[email protected]> wrote: > Hi Qin, > > Thank you for your summary. > > I prefer option (2) for two reasons. > > Firstly, there are two CellLists in the Request message (one of them is > optional), thus it is logical to have two CellLists in the Response > message. If the relocation verification fails, the second CellList is > elided. > > Secondly, if the option (1) is used, the size of the CellList is always > the same as the R.CellList in the Request message, therefore the relocation > failure is identified by a CellList of (NaN,NaN) cells; if the option (2) > is used, the failure is identified by an empty CellList, which maintains > the consistency of design philosophy with ADD and DELETE. > > Best regards, > Remy > > > From: Qin Wang [mailto:[email protected]] > Sent: Tuesday, September 12, 2017 4:37 AM > To: Liubing (Remy) <[email protected]>; Xavi Vilajosana Guillen < > [email protected]> > Cc: [email protected] > Subject: Re: [6tisch] Question about the Relocation in 6P > > Hi Remy, Xavi and all, > > Per the WG meeting on last Friday, we are going to solve the Relocation > issue before publishing new version of 6P draft. I put previous discussion > on the issue into three options as follows . Can we make choice from the > following options? > > (1) Introduce (NaN,NaN) as a cell value, change the definition of CellList > in the Response message, i.e. a list of new locations for all of the cells > which are required to be relocated. The cell value could be a cell in the > Candidate CellList in the Request message, or (NaN, NaN). > > (2) Two CellLists in the Response message. The first one indicates the > cells which can be relocated, and the second one indicates the related new > cells. If the number of elements in the first CellList is 0, it means the > relocation fails. > > (3) Keep what is in the current version > > Personally, I prefer (1). What do you think? > > Thanks > Qin > > > > On Thursday, September 7, 2017 3:50 AM, Liubing (Remy) < > [email protected]> wrote: > > Hi Qin and Xavi, > > Yes, that is what I meant. Thank you for your example. In this case, the > CellList in the response message will be the same size as the R.CellList. > This idea may bring more flexibility for the relocation SFs designed in the > future. Now I have another idea below. > > The current solution in 6P for ADD, DELETE, and RELOCATE maintains good > consistency in the design of the response message, i.e. the three > possibilities of the verification: succeed, fail, and partially succeed. > The three possibilities have the same return code set to SUCCESS, and are > distinguished by the number of the elements in the message. However, the > logic of RELOCATE seems to be more complicated compared to ADD and DELETE, > because there are two CellLists in the RELOCATE require message which > should have more possibilities. > > That is why I was thinking of introducing an empty cell (NaN, NaN) to > indicate a relocation failure of a specific cell. But it seems to be > inefficient to identify the result of the SF’s verification, because the > CellList in the response message will be the same size as the R.CellList in > the require message. And the complete relocation failure will be identified > by a CellList of empty cells. In order to do it more efficiently, the > return code can be changed to “FAILURE”, but it will break the consistency > with ADD and DELETE. > > I have another idea here: it can also have two CellLists in the response > message. The first one indicates the cells which can be relocated, and the > second one indicates the related new cells. If the number of elements in > the first CellList is 0, it means the relocation fails. And in this case, > the second CellList can be elided. In case of succeed and partially > succeed, the second CellList is mandatory. This idea can maintain the > consistency with ADD and DELETE. > > Which one do you think is better? Using the empty cell or using two > CellLists in the response message? > > Best regards, > > Remy > From: Qin Wang [mailto:[email protected]] > Sent: Thursday, September 07, 2017 3:39 AM > To: Liubing (Remy); Xavi Vilajosana Guillen > Cc: [email protected] > Subject: Re: [6tisch] Question about the Relocation in 6P > > Hi Ramy, > > I can see your point. I think, as you proposed, a list of relocated cells > in Response message may be a good idea, because it supports more > flexibility. Take Fig 15 as an example. Assume only (4,2) is available at > nodeB. If (1,2) is relocated to (4,2), then, the list of relocated cells in > Response message is [(4,2), (NaN, NaN)], otherwise [(NaN, NaN), (4,2)]. > Right? > > BTW, it is impossible for (1,2) and (2,2) to be of different type (Rx and > Tx) as you suggested, because all of the cells in both relocation list and > candidate list are under one CellOptions. > > Thanks > Qin > > On Monday, September 4, 2017 3:16 AM, Liubing (Remy) < > [email protected]> wrote: > > Hi Xavi, > > Thank you for your response. > > I think the cells are equivalent if they belong to the same bundle. For > example, the cells (1,2) and (2,2) are used together to transmit a > relatively large packet. In this case, the two cells should be considered > as a whole: if (1,2) cannot be relocated then (2,2) won't be able too; > otherwise, if (1,2) can be relocated and (2,2) can't, it might be > inappropriate to relocate (1,2) only, because it could cause packet loss. > > If (1,2) and (2,2) are of different purpose (to transmit different > packets) or of different type (RX and TX), they can be considered > independently: the relocation of (2,2) should be considered even if the > relocation of (1,2) fails. > > Indeed, the policy is implementation-specific, but it might be better for > 6top to support more possibilities. For example, a cell (NaN, NaN) could be > used to represent a relocation failure. > > Best regards, > > Remy > > From: Xavi Vilajosana Guillen [mailto:[email protected]] > Sent: Friday, September 01, 2017 5:18 PM > To: Liubing (Remy) > Cc: [email protected] > Subject: Re: [6tisch] Question about the Relocation in 6P > > Hi Remy, > > I think this can be an implementation decision. IMHO, when a node requests > a relocation like the one in Figure 15, it assumes that any of the > candidate cells is equivalent. This means that if [1,2] cannot be relocated > then [2,2] won't be able too. Seen in another way, the relocation may > happen in the list order consuming all possible candidate cells. This can > be seen as a policy that may depend on the implementation or SF rules so > other options may also be possible but are out of the scope of 6P.. > > Do you have a specific example where the case you present is relevant? > > regards, > Xavi > > 2017-08-30 8:29 GMT+02:00 Liubing (Remy) <[email protected]>: > Hello folks, > > I have a question about the relocation of cells in the draft > 6tisch-6top-protocol. > > In section 4.3.3, node A wants to relocate several cells and selects > candidate cells from its schedule for node B, then node B's SF verifies > which of the cells it can install in its schedule. The verification can be > partially succeed. If N < NumCells cells appear in the CellList, this means > first N cells in the Relocation CellList have been relocated, the remainder > have not. > > Does this mean that if the relocation of the first cells fails, there > would not be necessary to verify if the rest cells could be relocated? For > example, in Figure 15, if the cell (1,2) in the R. CellList cannot be > relocated to any of the cells in C.CellList, then (2,2) will not be > relocated even if it is possible to relocate it to (6,5)? > > Thanks, > > Remy > > _______________________________________________ > 6tisch mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/6tisch > > > > > -- > Dr. Xavier Vilajosana > Wireless Networks Lab > Internet Interdisciplinary Institute (IN3) > Professor > (+34) 646 633 681 > [email protected] > http://xvilajosana.org > http://wine.rdi.uoc.edu > Parc Mediterrani de la Tecnologia > Av Carl Friedrich Gauss 5, B3 Building > 08860 Castelldefels (Barcelona). Catalonia. Spain > > > _______________________________________________ > 6tisch mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/6tisch > > > _______________________________________________ > 6tisch mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/6tisch > -- _______________________________________ Thomas Watteyne, PhD Research Scientist & Innovator, Inria Sr Networking Design Eng, Linear Tech Founder & co-lead, UC Berkeley OpenWSN Co-chair, IETF 6TiSCH www.thomaswatteyne.com _______________________________________
_______________________________________________ 6tisch mailing list [email protected] https://www.ietf.org/mailman/listinfo/6tisch
