Hello Brian: > > Title : IPv6 Socket API for source address selection > > Author(s) : E. Nordmark, et al. > > Filename : draft-chakrabarti-ipv6-addrselect-api-03.txt > > Hi Samita, Erik, and Julien, > > Didn't see this sent to the IPv6 ML, but sending comments here, most are > editorial.
Thanks for the comment. Yes, we published the 3rd revision of IPv6 address selection API draft. I don't know why it did not get announced at the IPv6 ML. We intend to publish this document as an individual RFC. The document is under review by some of the IPv6 API authors. But any input from the wg is also appreciated very much by the authors. > ------------------------------------------------------------------------ > Overall comment: > > 1. The draft talks about setting flag X and flag not-X, but the only > place I see where this can happen is trying to set these together: > > IPV6_PREFER_SRC_CGA | IPV6_PREFER_SRC_NONCGA > AI_PREFER_SRC_CGA | AI_PREFER_SRC_NONCGA > > Am I missing something? Can I somehow turn-off the use of Care-of > Addresses as source addresses? Maybe it's implied that if I don't set > the flag in the setsockopt()/getaddrinfo() call that it turns-off using > that type of source address? > > ------------------------------------------------------------------------ The combination of flag X and not-X is not allowed - only one should be set at one time, becuase one packet has one source address. The default flags are described in section 7. For example, by default *_SRC_HOME is the choice of flag, ( it means COA address is not used by default as source address), but one can change that by using setsockopt() with *_SRC_COA. > Section 4: > > <netinet/in.h>. > > Not in6.h? > IPV6 Socket options are defined in netinet/in.h i,e. IPV6_HOPLIMIT, IPV6_NEXTHDR etc. <snip - editorial comments> Thanks for the careful review for editorial changes. We will review them to incorporate in the document. > IPv4-mapped IPv6 addresses, which are IPv4 addresses in a form that > can be used on an AF_INET6 socket, are supported in this API. In > some cases the IPv4-mapped addresses may not make much sense because > the attributes are IPv6 specific. For example, IPv6 temporary > addresses are not the same as private IPv4 addresses. However, the > IPv4 mapped-address support may be useful for mobile home address and > care-of-address. At this point it is not understood whether this API > has any value to IPv4 addresses or AF_INET family of sockets. > > The second to last sentence here is confusing (regarding mobile), can > you explain what you meant? It was discussed in the ipv6 list sometimes ago that IPV4-mapped IPv6 addresses may be useful for home-address and COA for IPv4 and IPV6 interoperability. It is not well understood, but I think it is possible that a MIPv4 compliant dual-stack node may want to comunicate with a IPv6 address. > ------------------------------------------------------------------------ > Section 10: > > Sometimes an application may have a requirement to only use address > /address/an address/ > > ------------------------------------------------------------------------ > The verification of temporary vs. public, home vs. care-of, CGA vs. > not, are performed by a new function defined for this purpose: > > #include <netinet/in.h> > > Not in6.h? > You mean ip6.h? ip6.h is mainly for advanced API. But this document is more of an extension to Base API. In that context, <arpa/inet.h> may be a choice. On the otherhand, all the address testing macros are in netinet/in.h. Good catch! Perhaps IPv6 Socket API authors can shine light on this. Thanks again, -Samita -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
