sks-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Sks-devel] [CONTRIB] sks.pod text for enabling IPv6


From: Phil Pennock
Subject: Re: [Sks-devel] [CONTRIB] sks.pod text for enabling IPv6
Date: Mon, 30 Mar 2009 21:45:47 -0700

On 2009-03-30 at 22:30 -0400, Yaron Minsky wrote:
> 2009/3/30 Phil Pennock <address@hidden>
> 
> > If I cared more for the O'Caml language than I do, I'd look into
> > contributing back patches to the language maintainers to solve this
> > problem.  But that doesn't solve the immediate problem, unless an SKS
> > release will be delayed pending a new language compiler release and time
> > for everyone to move across to it.
> >
> 
> For what it's worth, we could just wrap up the C functions using OCaml's
> FFI.  That's how I got bdb support in there.  I'm not saying it's
> necessarily worth the effort, but it is distinctly doable.  OCaml is a good
> UNIX citizen and allows one to use C functions without too much pain.

Yeah, we also discussed (earlier this month) O'Caml 3.11; that version
includes IPV6_V6ONLY support.  Is there conditional compilation
dependent upon the version of O'Caml which can be put into SKS (without
completely overhauling the build system)?

Since we're going with having to explicitly enable IPv6 the only
important knob left is for Linux systems and some others where IPv6
sockets accept IPv4 traffic by default, we need to be able to stop that;
the AI_ADDRCONFIG stuff can wait.

Do you want to mandate O'Caml 3.11 for SKS 1.1.1 ?  Otherwise, the C
call is:

  int on = 1;
  setsockopt(sock, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on));

If that can be invoked on the IPv6 reconsock then the membership tests
should, AIUI, work cleanly with :: in recon_addresses.

RFC 3493 section 5.3 describes IPV6_V6ONLY.


Myself, I use FreeBSD and I use languages from the Ports collection and
I avoid getting into manually building/maintaining major sub-systems
without Ports.  I suspect that if there's a reason that 3.11 isn't
making it into Ports then the Port maintainer for SKS can just patch out
the relevant call to Unix.setsockopt as being unnecessary (unless
someone has sysctl'd net.inet6.ip6.v6only to 0).

-Phil

Attachment: pgpIBqbrdL_Kj.pgp
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]