Hi all, - I guess using one context only means we plan to use one prefix most of the time (in Ras...). Is this very limitative? Are there usual scenarios using several? - Regarding implementation, contexts will probably come with Ras and be bound to prefixes. If you store these prefixes anyway in some prefix list, the context can be bound to the prefix list structures, hence the memory overhead is just one byte per new context. - The bit overhead in the HC encoding is 1, which seems acceptable, there are still 3 reserved bits.
Hence, in my opinion, if using one bit only creates issues, we should not hesitate to use two. Cheers, Julien -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Samita Chakrabarti Sent: jeudi 25 septembre 2008 00:07 To: Jonathan Hui Cc: 6lowpan Subject: Re: [6lowpan] HC Contexts - How Many? Hi Jonathan, 2 contexts looks good to me as well for simplicity, less memory usage. Thanks, -Samita On Tue, Sep 23, 2008 at 9:43 AM, Jonathan Hui <[EMAIL PROTECTED]> wrote: > > In moving the HC draft from -00 to -01, we added support for multiple > compression contexts so that more than one prefix (other than > link-local) can be compressed. > > The obvious question now is how many contexts are enough? It is clear > that two contexts are useful (one for the 6LoWPAN and another for the > common destination that all nodes may be sending to). But should we add more? > > Some thoughts: > - Fewer contexts allow the HC encoding to use fewer bits in > identifying the context in use. > - Specifying support for X contexts requires nodes to allocate enough > memory to maintain X contexts (ip addresses, timers, etc.). > - Additionally, we cannot simply say that we will change X to some > larger Y at some future date since the nodes that only support X will > not be able to support Y > X contexts. > - One argument for supporting more contexts is that it allows network > renumbering while allowing all nodes to communicate with compressed > addresses. My thought, however, is that renumbering of 6LoWPAN > networks should be rare and when they do occur it is okay to incur > some extra cost and communicate with full addresses during that time. > > So my thought is to keep the number of supported contexts small (2). > What do other people think? > > -- > Jonathan Hui > > > > _______________________________________________ > 6lowpan mailing list > 6lowpan@ietf.org > https://www.ietf.org/mailman/listinfo/6lowpan > _______________________________________________ 6lowpan mailing list 6lowpan@ietf.org https://www.ietf.org/mailman/listinfo/6lowpan _______________________________________________ 6lowpan mailing list 6lowpan@ietf.org https://www.ietf.org/mailman/listinfo/6lowpan
