Stuff that doesn't work yet in IPv6 patch

On Thu, Apr 03, 2003 at 04:47:45PM -0500, Tom Lane wrote:

The IPv6 patch seems to still be a few bricks shy of a load. Grepping
for places that handle AF_INET but not AF_INET6 revealed these
unimplemented features:

1. IDENT authorization. Fails if either local or remote address is IPv6.

I've actually written this already.

2. SSL. Postmaster allows SSL for AF_INET but not AF_INET6.

Hmm, it really shouldn't matter if it uses AF_INET or AF_INET6
... I should look into that.

3. Client address display in backend's ps display seems to be v4 only.

I'll take a look at that too.

4. pgstat code can only bind to 127.0.0.1 (v4 loopback). On a v6-only
machine this would not exist, would it?

I'm not sure, but I think I changed something about that. Can
you point me to that code?

Kurt

Kurt Roeckx <Q@ping.be> writes:

2. SSL. Postmaster allows SSL for AF_INET but not AF_INET6.

Hmm, it really shouldn't matter if it uses AF_INET or AF_INET6
... I should look into that.

Yeah, I suspect it just needs to replace the == AF_INET test with
an isAF_INETx() test. But I don't have the facilities here to
verify it.

4. pgstat code can only bind to 127.0.0.1 (v4 loopback). On a v6-only
machine this would not exist, would it?

I'm not sure, but I think I changed something about that. Can
you point me to that code?

src/backend/postmaster/pgstat.c

regards, tom lane

On Fri, Apr 04, 2003 at 06:37:17PM -0500, Tom Lane wrote:

4. pgstat code can only bind to 127.0.0.1 (v4 loopback). On a v6-only
machine this would not exist, would it?

I'm not sure, but I think I changed something about that. Can
you point me to that code?

src/backend/postmaster/pgstat.c

That was the code I was thinking it was. I changed the code
already.

I find it so weird, why isn't it just using socketpair() or
something?

Kurt

Kurt Roeckx <Q@ping.be> writes:

I find it so weird, why isn't it just using socketpair() or
something?

Portability issues, perhaps. Does socketpair exist and behave the same
everywhere?

Checking the HPUX man page for it, I read "socketpair() is supported
only for AF_UNIX", which suggests that it's not possible to get datagram
semantics from it, at least on this platform.

regards, tom lane

On Sat, Apr 05, 2003 at 12:12:57PM -0500, Tom Lane wrote:

Kurt Roeckx <Q@ping.be> writes:

I find it so weird, why isn't it just using socketpair() or
something?

Checking the HPUX man page for it, I read "socketpair() is supported
only for AF_UNIX", which suggests that it's not possible to get datagram
semantics from it, at least on this platform.

SUS says that only AF_UNIX should be supported too:

The socketpair() function is used primarily with UNIX domain
sockets and need not be supported for other domains.

That does not prevent you to set the type to either SOCK_STREAM
or SOCK_DGRAM, or even SOCK_SEQPACKET however.

Kurt

Kurt Roeckx <Q@ping.be> writes:

SUS says that only AF_UNIX should be supported too:
The socketpair() function is used primarily with UNIX domain
sockets and need not be supported for other domains.

Well, that just begs the question: does it even exist on platforms
that don't support AF_UNIX (eg, Windows, BeOS, ...)? If so, what
address family to use there?

regards, tom lane

Are these IPv6 deficiencies addressed?

---------------------------------------------------------------------------

Tom Lane wrote:

The IPv6 patch seems to still be a few bricks shy of a load. Grepping
for places that handle AF_INET but not AF_INET6 revealed these
unimplemented features:

1. IDENT authorization. Fails if either local or remote address is IPv6.

2. SSL. Postmaster allows SSL for AF_INET but not AF_INET6.

3. Client address display in backend's ps display seems to be v4 only.

4. pgstat code can only bind to 127.0.0.1 (v4 loopback). On a v6-only
machine this would not exist, would it?

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/docs/faqs/FAQ.html

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian wrote:

2. SSL. Postmaster allows SSL for AF_INET but not AF_INET6.

This is fixed and works now.

Regards,
Andreas

Tom, do you know how many of these issue are still open?

---------------------------------------------------------------------------

Tom Lane wrote:

The IPv6 patch seems to still be a few bricks shy of a load. Grepping
for places that handle AF_INET but not AF_INET6 revealed these
unimplemented features:

1. IDENT authorization. Fails if either local or remote address is IPv6.

2. SSL. Postmaster allows SSL for AF_INET but not AF_INET6.

3. Client address display in backend's ps display seems to be v4 only.

4. pgstat code can only bind to 127.0.0.1 (v4 loopback). On a v6-only
machine this would not exist, would it?

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/docs/faqs/FAQ.html

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian <pgman@candle.pha.pa.us> writes:

Tom, do you know how many of these issue are still open?

Uh, none of them, I would hope. That message was a long time ago.

regards, tom lane