Re: [CARD Techical Issue] Rate limiting flag
"Eunsoo Shim" <[email protected]>
| Newsgroups | gmane.ietf.seamoby |
|---|---|
| Message-ID | <003401c3416e$41eec180$ca6b0f8a@peace> |
> > As said earlier, the bit makes the protocol flexible without major > > additional complexity. > > The value of flexibility is worth adding the one-bit flag, I think. > > Also I guess the MNs would be need CARD information near handoff in most > > cases. > > That is, the CARD message traffic would be bursty to individual MN. > Setting > > a large inter-message time may not be desirable. Also allowing a too short > > inter-message time is not desirable, either. So the inter-message time > > should be configurable. Then the MN should be informed about the vale > which > > could be different at different ARs. This brings additional complexity, > > which the R flag can make unnecessary. > > So I think there are significant advantages with the R flag. > > > > I'm not arguing whether the inter-message time should be configurable, just > that a default should be included and that it should be configurable. > Well, we could recommend a default value that is configurable. But we wondered what would be the base in picking a certain number as the default value. > The problem here is that protocols which operate on the local link, like RFC > 2461, typically have explicit intermessage times, whereas end to end > protocols such as TCP (with ECN) do have explicit rate control. This > protocol is more like the former. > Well, even though we are more concerned about MN-AR interaction, the rate limiting should be applied to AR-AR interaction as well. The R flag will work for both cases. > I don't see that a rate flag makes the protocol any more or less flexible > than having an explicit intermessage time enforced by the host, if the > intermessage time is configurable. In either case, the protocol depends on > the host to enforce the rate control. I am repeating my self. If the inter-message interval is configurable, the MN should be informed about it. To do this, the AR should send a message to each MN, probably by periodic broadcast/multicast, or add additional fields to the CARD reply. Again more rf bandwidth usage in addition to more processing! The R flag removes this need. Isn't the R flag more flexible and simpler? Please let me ask a different question. What is the problem with the R flag you can see? Eunsoo