https://dnsdb-api.isc.org

keep output format and query format separated in two different documents
we should not specify which rtypes MUST be supported. It is only an output format spec
 rrtype should be mnemonic that is supported by IANA
 rfc 3597 unknown record type -> reference this rfc for arbitrary rrtypes
 same as for RDATA (if it can not format it correctly, it should be formated as in rfc3597

optional: x-more-data-coming: 50   (for Cisco)


we define the record format, any extensions should be specified in an RFC describing how the protocol works
--> specify the control packets in a separate RFC
--> Paul will ask Robert.
fearful of breaking existing programs by adding a bookend
Idea: if you can add &show_progress=1 to ReST GET request, then we can add the fields
If a given implementation supports bookends, progress bars, then these have to be optional and they should be signalled via HTTP GET &parameter

Extension to the record format must be x-* (for exaple x-query-id)

Samples are OK! But move them to a separate section!



dnstap -> get mail from paul
in google protocol bufs
will tap any part of a nameserver. 



next steps:
  * private circulation with robert edmonds, bfk, auckland, cert.ee

virustotal (?) -> later


submission:
  the dns-op still exists! So pdns-qof should be submitted to dns-op