[hydra-dev] requirement: ipv6 ?

Olle E. Johansson oej at edvina.net
Fri Jun 11 12:35:08 CDT 2010


11 jun 2010 kl. 19.03 skrev Simon Perreault:

> On 2010-06-11 12:20, Olle E. Johansson wrote:
>>> * Hydra MUST be capable of running in an ipv6-only network.
>>> * Hydra MUST be capable of running in an ip4-only network.
>>> * Hydra MUST be capable of running in a mixed ipv4 and ipv6 network.
>>> * in a mixed ipv4/v6 network, all symbolic addresses addresses MUST be resolved giving preference to ipv6.
>>> * Hydra must obey TTL for any cached addresses.
>>> 
>>> Am I missing any at this level?  is there agreement?
>> And we do need asynch dns from start as well as DNSsec support.
> 
> Be careful what you build into Hydra.
> 
> - Preference for IPv6: yes and no. If preference means "try IPv6 first,
> then IPv4", sure, no problem. However, this does not happen at the time
> of name resolution. The name resolution subsystem should return all
> results. Then when you actually try to make a connection, you have a
> list of addresses, sorted if you wish, that you try until one works.
> Bonus points if you try them simultaneously.
I think it should be configurable. One large vendor had hard-coded IPv4
preference since IPv6 was "only used over tunnels that are bad for media".

> 
> - DNS caching should be done separately, in a proper DNS cache, not
> within Hydra. For example, one could run Unbound (a lightweight caching
> nameserver) on the same host as Hydra. There should be no caching at all
> in Hydra itself. Otherwise you're just reinventing the wheel and
> inevitably introducing tons of bugs.
> 
> - DNSsec should be built into the OS, not re-done separately in each and
> every application. Hydra should not have to deal with that at all.
Yes, and no. There are reasons why I want apps to verify DNSsec in itself
and not rely on the outside and not know if a record is verified or not. 
Not for A/AAAA records, but other kinds of records.

> 
> - Async DNS: YES! ;)

Absolutely!

/O



More information about the asterisk-scf-dev mailing list