Victor Zhou wrote: > Dear DNSOP Working Group, > > I am pleased to submit the Internet-Draft "Prefixed TXT Records as Transition > Mechanism for New RR Types" (draft-zzn-dns-new-rr-00) for your consideration > and feedback.
Hi, I do not understand the proposed chunking mechanism. All resource records having the same owner name, class, and type are part of a resource record set and must be transferred atomically. Your proposed "chunks" are resource records that form an RRset, potentially together with other unrelated resource records that are not part of the chunked data. These "chunks" cannot be transferred independently. These quotes from your document do not make sense to me if the "chunks" are some or all of the RRs in a particular RRset: - "Record authors SHOULD keep chunks small enough to avoid UDP fragmentation (typically under 512 bytes including DNS overhead)" - "Implementations SHOULD implement a timeout mechanism for reassembly to handle cases where some chunks might be missing or delayed" - "The time spent waiting for missing chunks" - "Zone transfer mechanisms SHOULD ensure all chunks are transferred together to maintain consistency." Because all of the "chunks" will be received together or not at all. Either there is a new transport between resolvers and authoritative servers which allows retrieval of individual RRs within an RRset which is not described anywhere in the document, or the "chunks" need a scheme allowing them to be divided into multiple RRsets that can be retrieved in independent queries. Separately, I also agree with John Levine's comments regarding RFCs 8552, 8553, and the underscore registry. -- Robert Edmonds _______________________________________________ DNSOP mailing list -- dnsop@ietf.org To unsubscribe send an email to dnsop-le...@ietf.org
