Subject: Re: send-pr IPv6 patches?
To: None <>
From: None <>
List: tech-net
Date: 12/03/1999 00:19:48
	I've cc'ed to tech-net and set a reply-to.

>> It is unclear that we'll be able to port all programs to use
>> IPV6 in time for 1.5. It is a good thing to consider, and I
>> would support creating a PR category for it.
>I think most of the userland work that needs to be done is relatively
>simple work once you know the new socket APIs, so in principle a 
>lot of people could make useful contributions here (I mean you
>don't have to be one of the elite-kernel-hackers for this);
>provide patches and/or help build up a ToDo list. I think it
>would be nice to be able to advertise NetBSD-1.5 as the first
>OS with full support for IPv6.

	I would like to supply a list of "help needed" domain.
	- NFS.  INRIA IPv6 has some code, but need to be careful about
	  interop.  it seems Sun is working on it (including extending RPC).
	  I'm not sure if it is included in Solaris 8.
	- NIS.  For example, Solaris 8 seems to use{name,addr} for
	  IPv4-only and{name,addr} for dual-stack mappings.
	  should be trivial but need coordination/interop (no RFC/i-d yet).
	- X11.  There's a codebase (again INRIA), need some discussion on
	  interop issue, someone needs to merge the code back into XFree86
	  tree for ease of maintenance.
	- Kerberos maybe?  Heimdal has IPv6 support, I have no real chance
	  to test them.

	For porting, please be VERY sure to use get{addr,name}info,
	not getipnodeby{name,addr} nor gethostbyname2.  Scoped IPv6 address
	(like link-local and site-local) needs to be disambiguated when
	your machine is on site boundary, and for that we MUST use sockaddr_in6
	not in6_addr (there's a big debate on ipngwg ongoing).
	Basically we need to remove AF dependency from application code.
	Here are some documents: