Michel: On Sat, 15 Feb 2003, Michel Py wrote:
> Alan, > > > Alan E. Beard wrote: > > No wonder we're at an impasse: we have a blind-men-and-the-elephant > > problem here! > > No. We know what the elephant looks like. This is not the issue. The > issue is that for the last eight years we have been trying to design an > animal that carries as much as an elephant, at the speed of a cheetah, > drinks as little as a camel, an possibly poops only in the toilet and > flushes when done. > In the context of the overall architecture and design of the protocol, you are, of course, right. It seems a bit hyperbolic to suggest that we don't have a reasonable conception of the overall architecture of IPv6. The blind-men-and-the-elephant metaphor was advanced in the context of the address allocation practices discussion. It does seem, at least to this observer, that we have a number of groups working on discrete issues which, in aggregate, may arise from a common root cause, and that an acknowlegment of that common root cause has so far been, at best, tacit. > > > The common, underlying issue, as I see it, is: > > The use of PA space in end-user networks has the effect of imposing > > upon those networks functional burdens and restrictions in multiple > > areas which the managers of commercial end-user networks may be > > unwilling to tolerate. In consequence, we may need to reconsider > > our current address allocation practice, which relies principally > > on the PA model, in light of current user expectations, current > > state of the routing protocols and standards, and anticipated > > developments in routing and switching code. > > Given the rest of your postings, this appears to be a rather > black-and-white view from the enterprise point of view. For the same > reason multi6 has failed by limiting the scope to site multihoming, this > approach is equally doomed to fail because it ignores the issues of > large operators. > This criticism would be both valid and compelling if the proposal cited above had been worded more narrowly than the text shows. In fact, the call for reconsideration does not ignore the interests of the large operators: the specific text is :"in light of current user expectations". Last time I looked, the class "user" subsumed the class of large operators, as well as privately operated networks (commercial and otherwise), academic networks, retail ISPs, and others. Granted, the problem statement does include a specific reference to commercial operators of private networks. We have been dealing with a number of technical issues which have adverse impact on that class of network. However, the deleterious effects are not limited to privately-operated commercial end-user networks. Perhaps a broader problem statement is in order, but the fundamental issue remains. If the call for reconsideration had been worded to exclude any one or more of the classes of network users, or to be so narrowly specific as to admit _only_ the interests of selected classes of network users, the criticism of "black-and-white view from the enterprise point of view" would be irrefutable; however, such is not the case, and the criticism fails on that ground. > There is some quality work that has been done in the field of PA > multihoming solutions as well, such as: > http://www.ietf.org/internet-drafts/draft-huitema-multi6-hosts-01.txt > There is no suggestion (at least to my memory) that the work cited above, and others of like character, should not be part of the solution space. The specific work cited would be rendered inapplicable only if use of PA space were abandoned entirely for all or most classes of users; since no proposal to abandon PA exists (nor is any such proposal anticipated - more on this below), we should reasonably expect work such as that cited above to continue and, wherever shown technically sound, to see actual service in production networks. The general tenor of the comments above (and to some extent, of that below) appears to proceed from an assumption that the writer of the original call for discussion (and, before the question is raised, I did indeed write the text of the call for discussion) presupposes an outcome similar to: "abandon PA (largely or entirely); use PI as the preferred addressing model; and leave the transport and service providers to deal with the consequences, whatever they may be". I can state from direct and authoritative knowledge that the writer of the call for discussion presupposes no particular outcome whatsoever. Furthermore, the writer of the text of the call for discussion would be very likely to oppose any outcome of sort suggested above as signally irresponsible, and unworkable in practice, given the present state of our technologies. Given the way the IESG, the IETF, and this working group opreate, it would be grossly presumptous of any person to presuppose any particular outcome, or even to hold expectations concerning the character or the outline of any such outcome. > Although the current IPv6 architecture indeed favors the large operator > over the enterprise to the point that the enterprise network manager's > feet feel the IPv6 water too cold to put more than the tip of the toe in > (and this needs more balance), the other side of this coin is that if > one wants to make a buck out of IPv6 today one has to look at Asia and > to a lesser extent Europe for markets that involves a large percentage > of mobile devices that are a good fit for the PA model. > The statement immediately above seems a good summary of the general state of our current address allocation philosophy and the technical issues attendant thereunto. Yes, it would appear that PA space is probably appropriate for several classes of use, and that use of PA space probably should not be abandoned. However, as pointed out above, more balance is indeed needed with regard to address allocation policy for some classes of users: it may not be desirable to continue to prefer PA allocations (to the nearly total exclusion of other models) for at least some user groups. Let's talk the matter over, shall we? > No bucks, no Buck Rogers. > > Michel. > Personal note: Michel, your last paragraph above indicates, at least by my interpretation, that you and I are really after the same objective: a technically sound and balanced policy which will faciltate the widest possible acceptance and implementation IPv6 in both pubic and private networks. Thank you for your comments, and for providing this opportunity to correct any misapprehensions which may have arisen concerning the nature and intent of the call for discussion. Regards, Alan -- Alan E. Beard <[EMAIL PROTECTED]> AEBeard Consulting; 4109 Chelsa Ln; Lakeland FL 33809 863.815.2529 -------------------------------------------------------------------- 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] --------------------------------------------------------------------