[getdns-api] OPT Record fields

Willem Toorop willem at nlnetlabs.nl
Tue Feb 18 01:53:28 MST 2014


I prefer dropping the "class" and "ttl" keys and replace them with the
OPT specific keys.  I dislike having both "class" and
"udp_payload_size", and "ttl" and the other options because it is then
possible for them to have conflicting values.

We already have the same problem with "rdata_raw" and the split out
rdata fields.  Currently our implementation chooses "rdata_raw" above
the rdata fields when converting to wire-format, for example for DNSSEC
validation.

-- Willem

op 14-02-14 17:20, Goyal, Neel schreef:
> We could always keep ttl and class in the definition even though they
> might be Œjibberish'.  Then the spec can be amended to include the fields
> below.  Perhaps that¹s a reasonable compromise?
> 
> On 2/14/14, 11:18 AM, "Paul Hoffman" <paul.hoffman at vpnc.org> wrote:
> 
>> On Feb 13, 2014, at 1:26 PM, Goyal, Neel <ngoyal at verisign.com> wrote:
>>
>>> Section 4.1 of the spec states that Resource Records have the following
>>> fields: type, class, ttl, and rdata.  The OPT record is one of those
>>> weird ones where class is interpreted as a udp payload size, and ttl is
>>> actually broken up into extended rcode, version, z, and do bit.
>>>
>>> I propose we call out that the OPT RR will not have ttl and class, and
>>> instead be replaced with the following fields:
>>> 	€ udp_payload_size
>>> 	€ extended_rcode
>>> 	€ version
>>> 	€ do
>>> 	€ z
>>> type, rdata_raw, and options are already called out in the spec and are
>>> valid fields.
>>>
>>> Thoughts?
>>
>> I'm always of two minds on this one. OPT is an uglyandbrilliant hack, and
>> we should keep the names the same as for other RRs. Or: people who don't
>> know OPT need to be reminded with the new fields.
>>
>> So, I'll defer to others.
>>
>> --Paul Hoffman
> 
> 
> _______________________________________________
> getdns-api mailing list
> getdns-api at vpnc.org
> 



More information about the getdns-api mailing list