Hi,
Since installing Parallels 6 on my Mac, I've been having strange networking issues. Today I finally had the time to investigate. I wish I hadn't. It's all quite a mess:
Parallels version: Build 6.0.11994
OSX version: 10.6.7
System status: Parallels NOT running
1. Router advertisements
Somehow Parallels thinks it's a good idea to send IPv6 Router Advertisements on vnic0. This means that there'll be an IPv6 default route pointing to vnic0 on the OSX host:
This messes with IPv6 connectivity. Even if you then disable IPv6 on your primary ethernet interface, the host system will still assume there is IPv6 connectivity, causing lengthy timeouts for anything that has an AAAA record in the DNS.
2. Depricated addresses
Parallels configures the IPv6 addresses ::0.0.0.1 and ::0.0.0.2 on vnic0 and vnic1. These are IPv4-compatible-IPv6 addresses and have been depricated in RFC 4191 way back in 2006. Even *if* the developers have been living under a stone since then, RFC3513 demands:
0.0.0.1 and 0.0.0.2 can of course hardly be called "globally-unique IPv4 unicast addresses"
3. While digging around I found the network.desktop.xml file. A nice mess there as well:
- Six of the eight defined interfaces there have a VLAN tag of 65535. I don't know if that's meant seriously, as VLAN tags are only 12 bits, so 0-4095, not 65535.
- While it's great that Parallels supports DHCPv6, I'm not sure what this is about:
- 127.0.0.1 isn't a very nice IPv6 address. I also supect that network is already in use on another interface.
- A little further down we find:
Why is the loopback address the DHCP scope?
While it is commendable that Parallels tries to build IPv6 support into its software, this design and implementation could still use a bit of work. With this setup, my OSX Host's IPv6 connectivity is broken, even with Parallels NOT running. As APNIC has already run out of its available supply of unallocated IPv4 space and the other RIRs will follow shortly, this issue could use a bit of attention.
Best regards,
Alex
Since installing Parallels 6 on my Mac, I've been having strange networking issues. Today I finally had the time to investigate. I wish I hadn't. It's all quite a mess:
Parallels version: Build 6.0.11994
OSX version: 10.6.7
System status: Parallels NOT running
1. Router advertisements
Somehow Parallels thinks it's a good idea to send IPv6 Router Advertisements on vnic0. This means that there'll be an IPv6 default route pointing to vnic0 on the OSX host:
This messes with IPv6 connectivity. Even if you then disable IPv6 on your primary ethernet interface, the host system will still assume there is IPv6 connectivity, causing lengthy timeouts for anything that has an AAAA record in the DNS.
2. Depricated addresses
Parallels configures the IPv6 addresses ::0.0.0.1 and ::0.0.0.2 on vnic0 and vnic1. These are IPv4-compatible-IPv6 addresses and have been depricated in RFC 4191 way back in 2006. Even *if* the developers have been living under a stone since then, RFC3513 demands:
0.0.0.1 and 0.0.0.2 can of course hardly be called "globally-unique IPv4 unicast addresses"
3. While digging around I found the network.desktop.xml file. A nice mess there as well:
- Six of the eight defined interfaces there have a VLAN tag of 65535. I don't know if that's meant seriously, as VLAN tags are only 12 bits, so 0-4095, not 65535.
- While it's great that Parallels supports DHCPv6, I'm not sure what this is about:
- 127.0.0.1 isn't a very nice IPv6 address. I also supect that network is already in use on another interface.
- A little further down we find:
Why is the loopback address the DHCP scope?
While it is commendable that Parallels tries to build IPv6 support into its software, this design and implementation could still use a bit of work. With this setup, my OSX Host's IPv6 connectivity is broken, even with Parallels NOT running. As APNIC has already run out of its available supply of unallocated IPv4 space and the other RIRs will follow shortly, this issue could use a bit of attention.
Best regards,
Alex
