[bind10-dev] Convention for address+port: forget rough consenus, lets vote!

[Tomasz Mrugalski]

On 11-08-10 21:42, Shane Kerr wrote:
> All,
> 
> We had a thread last month about how to present address+ports in BIND
> 10:
> 
> https://lists.isc.org/pipermail/bind10-dev/2011-July/002459.html
> 
> We don't think any consensus was reached. So, we're going to try an
> alternate strategy: voting.
> 
> With the help of the some of the team, I created a doodle poll:
> 
> http://www.doodle.com/rtbp6eb3xpmxre8q
> 
> Please let us know what you think!
> 
> Everyone is welcome to vote - this is just a poll to get an idea of what
> general opinion is.
Can we also add a clarification for interface name? In DHCPv6
address/port is not enough. There are link-scoped addresses that are the
same for each interface, e.g. [ff02::1:2]:547.

Of course interface name should be used only when it is really
necessary. Assuming that currently leading option (Option 3 in doodle
poll: IETF standards (RFC 2396 and RFC 2732): v4address:port and
[v6address]:port) wins, we could use one of the following:

1. [ff02::1:2%eth0]:547
2. [ff02::1:2]:547%eth0
3. [ff02::1:2]:547 at eth0
4. [ff02::1:2 at eth0]:547
5. eth0/[ff02::1:2]:547

1. and 2. are based on notation used by OS X (and most BSDs, I think).

I've seen @eth0 notation somewhere, but I can't remember where it was.

5. Is somewhat similar to what is used in ISC DHCP. ISC DHCP uses
interface-name/mac-address/address/prefix format, but it is somewhat
inconsistent in various places. Example:

eth0/00:01:02:03:04:05/192.168.0.2/24

I think the most reasonable option is 1, but if there are objections,
maybe that could be a subject for another poll once the current one is
finished?

Tomek