Keith, I agree yet again on your scaling well put statements below. I thought of that and between the lines when I said "how does the user know which interface" I was really addressing the scalability issue. But awhile ago I came to realize that the only time that argument works is when it is used as filerbuster to hold up engineers work in the IETF and I have seen it used on a few specs. A good example I gave up on was router preferences for IPv6 I argued we had to have them but got shot down based on ones reputation. I think everyone one of us do router preferences with ND in implementation as Hosts today and just got it done. We could have done so much more with ND had we looked at router preferences but all at that time in the IPv6 world were anti-hosts knowing anything about IPv6 and I believe anti-servers for operations (e.g. DHCPv6, Service Location) and why we still have not completed DNS Discovery in the IETF. So I gave up on using scalable hypothesis in debate. What I tried here was simple common sense with some network diagrams and such. Then folks started screaming "but Jim this ipv6 advantage". Of course it is but not for applications use. But LLs will not scale.
P.S. the one that still drives me nuts is when someone tells me I have a packet built in my kernel where the source has less scope than the destination address, just send it anyway? Don't get it and never will :--). /jim > -----Original Message----- > From: Keith Moore [mailto:[EMAIL PROTECTED] > Sent: Friday, August 22, 2003 12:02 AM > To: Bound, Jim > Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED]; > [EMAIL PROTECTED]; [EMAIL PROTECTED] > Subject: Re: Some IPv6LL operational experience > > > On Thu, 21 Aug 2003 23:44:37 -0400 > "Bound, Jim" <[EMAIL PROTECTED]> wrote: > > > I came here and asked a simple question. Look what happened. In > > industry this would be a no brainer and most would take > your view is > > my opinion. Don't use these for applications. But it has > turned into > > another absurd avoidance of basic principles because folks > have their > > heals dug in on for some reason. > > As far as I can tell, the "reason" is a basic failure to do > engineering, or more specifically analysis. A small number > of emperical tests with limited scope are taken as evidence > that some practice or another will not cause problems for an > extremely diverse Internet with 10**8 - 10**9 distinct users > and 10**5 - 10**6 different apps. That's the networking > equivalent of driving a single car over the original Tacoma > Narrows Bridge and declaring it sound. > > (I cite the Tacoma Narrows Bridge because it was another case > of ignoring the analysis that was done (of wind effects) - > blind faith in the designer's reputation was deemed to make > such analysis irrelevant.) > > Keith > > p.s. If you're not familiar with that bridge, > see http://www.camerashoptacoma.com/mpegs/TacomaNarrowsBridge.mpg > -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
