Tengfei, So, taking into account your proposal, the SF should configure the Retry Limit and handle the retries at SF layer, leaving 6P to deal only with transactions: either they are complete and successful or failed and rolled back (even in a timeout case). Retries shall not be allowed during a 6P transaction: This increases delay, but reduces complexity. What do you think? Regards,
Diego 2016-03-04 6:00 GMT-03:00 Tengfei Chang <tengfei.ch...@gmail.com>: > Hello Diego, > > *For IANA_6TOP_CMD_ROLLBACK command: * > > I think I understand what you are trying to solve. But using > IANA_6TOP_CMD_ROLLBACK > has another issue: how to decide when to send the command. I think this is > the same question for setting the value of 6P_TIMEOUT. > Also if a transaction timeout because of bad link quality or lost > connection, sending another packet is not a good idea. > > *For Retry handling:* > > First, I think whether retry or not is a decision made by scheduling > function. For 6P, it just forwards this message to upper layer (such as SF0) > > If retries are required, I think it makes sense that retrying within a > transaction. > > What do you think? > > Tengfei > > On Thu, Mar 3, 2016 at 9:41 PM, Prof. Diego Dujovne < > diego.dujo...@mail.udp.cl> wrote: > >> Dear all, >> One of the sections I would like to add to >> the SF0 draft is the examples of error handling. >> >> One example could be Timeout handling: >> When a transaction times out, either a packet is lost, >> the destination node lost connectivity or it is too busy >> to answer on time. >> Consequence: The transaction, in any stage, should >> be rolled back by both ends. >> Here, both nodes are expected to Timeout. >> Section 3.6.2 of the sublayer draft "Aborting a >> 6P transaction", cannot be used, since in this case >> there is no response. >> I propose to issue a rollback command such as >> "IANA_6TOP_CMD_ROLLBACK" to the destination >> node using the same token from the original transaction, >> and do not wait for an answer. >> This enables the source node to try a new transaction >> before the destination node times out, thus reducing >> delay. >> >> Would you agree in adding this command to reduce >> delay? >> >> Another issue should be Retry handling, in case of >> timeout. >> Do we allow retries within a transaction? >> If we allow retries, the maximum number of retries before >> considering a transaction failed should be configured on >> the SF. >> >> Thank you. >> Regards, >> >> Diego Dujovne >> >> >> -- >> DIEGO DUJOVNE >> Académico Escuela de Ingeniería en Informática y Telecomunicaciones >> Facultad de Ingeniería UDP >> www.ingenieria.udp.cl >> (56 2) 676 8125 >> >> _______________________________________________ >> 6tisch mailing list >> 6tisch@ietf.org >> https://www.ietf.org/mailman/listinfo/6tisch >> >> > -- DIEGO DUJOVNE Académico Escuela de Ingeniería en Informática y Telecomunicaciones Facultad de Ingeniería UDP www.ingenieria.udp.cl (56 2) 676 8125
_______________________________________________ 6tisch mailing list 6tisch@ietf.org https://www.ietf.org/mailman/listinfo/6tisch