In message <1AE7298D-CC55-4284-AA3F-7492249C4182@xxxxxxxxx>, =?utf-8?Q?Patrik_F =C3=A4ltstr=C3=B6m?= writes: > > On 27 feb 2015, at 16:02, Andrew Newton <andy@xxxxxx> wrote: > > > > - I am also nervous over the size of the RRSet, i.e. same issue I see > with NAPTR, and the reason why I added the prefix (like SRV) to the URI RR > > > > I plan to move to use this advice. Thanks. > > The encoding is important, I really like having multiple TXT strings > (length-prefixed) but then "the last one" should be "the rest of RDATA" > > Maybe Mark and others should think about how to do that encoding of the > RDATA... I am lacking ideas here. Part from telling how many TXT strings > there are. > > paf I don't really care about the encoding as long as it is *stable*. The wire format has gone from a series of <len><value> pairs that get concatenated for use (up to draft-faltstrom-uri-07) to just <value> with a implict length (draft-faltstrom-uri-08). The type code was allocated against draft-faltstrom-uri-05. Below are some of the tradeoffs that result from the different wire encodings. Master files need to preserve wire encoding (for DNSSEC) so <len><value> pairs restrict how master files are written. This leads to lines which are potentially over a 1000 bytes long due to \DDD encodings of non-ascii. With a plain <value> and concatenation you don't have to have long lines in master files and you can print the field out over multiple lines knowing you can recombine the field. example.net. URI ( 0 1 "http://foo.bar/"; "xxx?lhllshdflldhlh;" "yyy?lhjjgjfkakfjkfjak" ) Mark -- Mark Andrews, ISC 1 Seymour St., Dundas Valley, NSW 2117, Australia PHONE: +61 2 9871 4742 INTERNET: marka@xxxxxxx