on 2005-08-31 05:40 Jeffrey Hutzelman said the following: > > On Tuesday, August 30, 2005 15:55:56 -0700 Ned Freed > <ned.freed@xxxxxxxxxxx> wrote: > >> IMO this needs major work even before being approved as experimental. The >> overlapped namespace approach in particular seems hugely problematic and >> IMO needs to be replaced. > > I've only read this document briefly, but based on what I've seen and on > the descriptions and explanations in the current discussion, I have to > agree. The overlapped namespace approach has significant problems, which > have been mentioned here. It generates load in the form of additional > queries on caching servers and on the global DNS roots for names those > servers are never expected to be able to resolve, and in the form of > multicast traffic on the local link for potentially every failed query > against the global DNS. > > > It also creates massive ambiguities in the namespace, by allowing any host > on the local link to claim any global DNS name which happens not to resolve > at the moment (even if due to a temporary failure). This means that names > which are intended to be part of the global DNS namespace may resolve > differently depending on one's location, or what hosts might be responding > to LLMNR requests on the local network. > > This is a problem so egregious that the IAB wrote a document about it > (RFC2826). While the majority of that document pertains specifically to > recurring "alternate root" proposals, much of it applies equally well here > -- "alternate roots" are a bad idea because they split what needs to be a > single global namespace into several alternate namespaces. The use of the > overlapped-namespace approach with LLMNR does the same thing, only instead > of creating a few alternate roots, it creates millions. Good summaries, good points. I do not believe the LLMNR specification should be published in its current form; the namespace confluence is extremely bothersome, and should not be accepted even for publication as an experimental RFC. Even if the namespace confluence problem is corrected, it seems more appropriate - given the deployment of mDNS - to publish both mDNS and LLMNR as experimental RFCs. Henrik. _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf