----- Original Message ----- From: "Ignatios Souvatzis" <[email protected]> Cc: <[email protected]> Sent: Thursday, September 22, 2011 9:34 AM > On Tue, Sep 20, 2011 at 09:34:46AM -0700, Stig Venaas wrote: > > On 9/20/2011 2:03 AM, Iljitsch van Beijnum wrote: > > >On 20 sep 2011, at 9:58, [email protected] wrote: > > > > > >>For pt2pt SDH links you want /127 to avoid the ping-pong problem, > > >>not /126. > > > > > >That's nice (as long as your routers ignore the all zeros anycast address), but the question was: how do we write down IPv6 prefixes? Are we ok with simply leaving off the undefined part, or do we insist on proper inclusion of the zeros? If the former, how do we avoid ambiguity for prefixes such as ::ffff/96? > > > > I agree it is ambiguous in some sense. That is, I expect any reasonable > > implementation to treat ::ffff/96 as ending in ffff, and being > > equivalent to ::/96. > > mostly... > > > However, for us humans, why would you write ::ffff/96 and not ::/96? > > Because I want to specify the address of an interface and its on-link > network space at the same time.
Which I think a mistake. They are two separate pieces of information and there is no standard way of writing that, as this thread has identified. RFC5952 has been cited as settling the question, but it does not. That RFC specifies a 'IPv6 Address Text Representation', it does not specify a notation for prefixes, or for prefixes and interface addresses combined in one token. '/' mostly appears in that RFC as part of a URL, with only a brief mention of its use in 'whois', never as part of the formal specification, so prefixes just are not covered there. Nor do I recall any discussion in the lead up to that RFC of how to specify a prefix or prefix/address combination. So if I saw ::ffff/96, I would assume that the writer did not fully understand IPv6 addresses and would want to clarify the matter before placing any reliance on it. IPv4 is different. We don't have the long strings of zeroes that could appear anywhere, we know that if the representation is incomplete, then it is the most significant part that is present, the least significant part that must be supplied. Tom Petch > > I think one can pretty much expect that the bits beyond the prefix length > > are 0 when writing prefixes. > > I don't see a big deal with using ::ffff/96 as a sloppy notation on > > whiteboards etc. > > For the reason I gave, I don't think so. I don't want a sloppy notation > to be in use to force me to context-sensitive parsing of addresses; this > will lead to mistakes, when the context is lost. Especially probable with > whiteboard notation which will copied manually, maybe only partially... > > > But for documents like Internet Drafts and RFCs, we > > should probably make sure to write the trailing 0s. > > With this I agree. > > > I myself having got some pushback for writing e.g. 239/8. Even though > > it is not ambiguous. > > Well, but the shortened notation for IPv4 networks has its roots in the > classful and pre-classful IPv4 notation, and is justified from there at > least to some degree... we don't have that... luxury... with v6. On the > other hand, we did specify a syntax for shortening (compression of zeros) > for IPv6, so let's stick to that, even in our whiteboard notes. > > Regards, > -is -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
