Hi, "node ETX" means the end-to-end path ETX, sum of ETX of all constituent link, from a node to the root, not the ETX of the local link to the node. When the white bit is set, the local ETX to the new node is regarded as 1, not the end-to-end path ETX. Hope this helps clarify my question. Thanks.
On Thu, Aug 12, 2010 at 1:25 AM, Manjunath Doddavenkatappa < [email protected]> wrote: > > In the terminology you defined, > > "node ETX" is what white bit indicates. I.e., Set white bit indicates that > "node ETX" of the new link is smaller (kind of ETX=1). As "node ETX" of the > evicted node can not be less than 1 and its "ETX" (to the Root) is greater > than that of the new node, it is OK to evict the node. > > > Regards, > Manjunath D > > > ################################################################################################################ > > *************************************************************************************************************** > > > On Wed, 11 Aug 2010, Xiaohui Liu wrote: > > Hi, >> >> Thanks for your reply. >> >> I guess there is some misunderstanding here. When I say ETX, I mean ETX >> from >> the neighbor to the root, not the ETX of the local link to the neighbor. >> Let's call this ETX the node ETX for ease of exposition. If the network >> layer can find some existing neighbor in neighbor table with worse node >> ETX >> than the new neighbor and recommends insertion, why the underlying >> estimator >> simply evicts a random neighbor regardless of its node ETX? This can evict >> a >> neighbor with smaller node ETX. Look forward to your further explanation. >> >> On Wed, Aug 11, 2010 at 12:39 PM, Philip Levis <[email protected]> >> wrote: >> >> >>> On Aug 10, 2010, at 10:50 PM, Manjunath Doddavenkatappa wrote: >>> >>> >>>> Just a guess, >>>> >>>> Before asking the routing layer whether a new route to a neighbor is >>>> promising, the estimator >>>> asks physical layer whether the white bit of the incoming packet (from >>>> >>> the >>> >>>> sender of the new link) is set. Only if the white bit is set then >>>> estimator proceeds. Since the set white bit already indicates that local >>>> link is good (may be interpreted as ETX=1), it may not be required to >>>> >>> verify >>> >>>> the local ETX values of the existing neighbors. >>>> >>>> Please correct me if I am wrong. >>>> >>>> Manjunath D >>>> >>> >>> Sort of -- please refer to the 4-bit link estimation paper. >>> >>> Normally, when the estimator first learns about a neighbor, it waits >>> before >>> making communication with that neighbor available (actually putting it >>> into >>> the link table as an active link). The reason is simple: after receiving >>> only one packet, the estimator can't provide a good estimate, and so >>> making >>> the link active might cause a protocol to choose a very very poor link. >>> >>> The white bit circumvents this initial estimation phase. The white bit >>> indicates that there's a high probability that the underlying link is >>> high >>> quality; this allows the link estimator to skip the initial estimation >>> and >>> make the link immediately available for the routing layer to use. >>> >>> Phil >>> >> >> >> >> >> -- >> -Xiaohui Liu >> >> -- -Xiaohui Liu
_______________________________________________ Tinyos-help mailing list [email protected] https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
