Repository : ssh://git@open-mesh.org/doc
On branches: backup-redmine/2019-09-17,master
commit c6762f87d67a877e098a5e2cac54c20cf2af8b71 Author: Sven Eckelmann sven@narfation.org Date: Tue Sep 17 10:29:16 2019 +0000
doc: batman-adv/RIP
c6762f87d67a877e098a5e2cac54c20cf2af8b71 batman-adv/RIP.textile | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-)
diff --git a/batman-adv/RIP.textile b/batman-adv/RIP.textile index 71e010f..73dd648 100644 --- a/batman-adv/RIP.textile +++ b/batman-adv/RIP.textile @@ -34,21 +34,21 @@ h2. Discussion
From the algorithm sketch, we can see:
- * we can promptly react to link outages - as soon as ELP (or some other underlying mechanisms) detects it. - * the death notes are broadcasted only through the affected part of the mesh. Outside nodes use Unicast to reach the Originator - this saves bandwidth - * we can keep long originator interval and still react to changes fast - at least negative ones. +* we can promptly react to link outages - as soon as ELP (or some other underlying mechanisms) detects it. +* the death notes are broadcasted only through the affected part of the mesh. Outside nodes use Unicast to reach the Originator - this saves bandwidth +* we can keep long originator interval and still react to changes fast - at least negative ones.
- * Remark from me/Linus: Deleting the routes (meaning forgetting about the last seqno+metric?) via the death notes can cause routing loops, if a late OGM arrives at N3 via yet another alternative path, I think. Alternative, safer suggestion: Mark entry as "undead", meaning to decrease the OGM_MAX_ORIG_DIFF for this specific entry to 1 until it gets updated. - * The name is awesome, fitting and easy to remember. However, there actually exists a prominent distance vector routing protocol with the same name. Should we change it, for the same reason we changed NDP to ELP? (if yes, a suggestion: BATSIG - BATMAN Sequence number Increment Generator; in reference to the bat signal which Gordon might use after looking at a dead, mutilated body) +* Remark from me/Linus: Deleting the routes (meaning forgetting about the last seqno+metric?) via the death notes can cause routing loops, if a late OGM arrives at N3 via yet another alternative path, I think. Alternative, safer suggestion: Mark entry as "undead", meaning to decrease the OGM_MAX_ORIG_DIFF for this specific entry to 1 until it gets updated. +* The name is awesome, fitting and easy to remember. However, there actually exists a prominent distance vector routing protocol with the same name. Should we change it, for the same reason we changed NDP to ELP? (if yes, a suggestion: BATSIG - BATMAN Sequence number Increment Generator; in reference to the bat signal which Gordon might use after looking at a dead, mutilated body)
h3. Death Note packet format
Death Notes and OGM Update Requests may share a packet type and be distinguished by a flag. They basicly contain:
- * the affected Originator Address (either used as subject of the death note or final destination) - * TTL - * (maybe?) last valid OGM +* the affected Originator Address (either used as subject of the death note or final destination) +* TTL +* (maybe?) last valid OGM
h3. Optimization Ideas
- * we may also send the last valid OGM from a node which is not affected by a death note - but only after making sure that the connection to the destination still works (e.g. by using a ping). This would require some more state to be saved. \ No newline at end of file +* we may also send the last valid OGM from a node which is not affected by a death note - but only after making sure that the connection to the destination still works (e.g. by using a ping). This would require some more state to be saved. \ No newline at end of file