Hi Elwyn,
I was not subscribed for NSIS_IMP mailing list. I have just done it. Xiaoming forwarded to me your discussion and I have some questions. Please see inline. ( i kept only the relevant topics)
<br />>At 13:14 09/02/2005, you wrote: <br />>Hi Elwyn <br />>I'm just trying to ask you a few question about the red comment: <br />> <br />>1. I think use of cookie (i.e., 3-way handshake) is orthogonal to <br />>piggyingback NSLP in Query. I believe the spec must say piggyback in query <br />>is only used when no cookie is used. <br /> <br />>That could well be a better idea... as defined at present *every* query has <br />>a Query-Cookie even if MRS state is not being installed, and one can <br />>piggyback on any Dmode setup message - as you say they are orthogonal. I <br />>think the whole issue of distinction between MRS to be installed and MRS <br />>not to be installed is very unclear... Tseno's FSM diagrams come up with a <br />>different idea from the one i deduced from the text.
What do you mean by "MRS state is not being installed"? I think that a MRS state installation follows every successful QUERY/RESPONSE/(possible)CONFIRM message exchange?
There is also this
<br />>major problem with policy conflicts between nodes... if some nodes won't do <br />>anything that might affect state (like sending things to the NSLP) before <br />>confirm, but others will, how is the NSLPdata to be handled? when is it to <br />>be transmitted?
Greetings, Tseno_
nsis_imp@informatik.uni-goettingen.de