- From: Nico Williams <nico@cryptonector.com>
- Date: Mon, 2 Dec 2013 20:23:06 -0600
- To: John Cowan <cowan@mercury.ccil.org>
- Cc: Phillip Hallam-Baker <hallam@gmail.com>, Tim Berners-Lee <timbl@w3.org>, IETF Discussion <ietf@ietf.org>, JSON WG <json@ietf.org>, Alex Russell <slightlyoff@google.com>, Tim Bray <tbray@textuality.com>, "Matt Miller (mamille2)" <mamille2@cisco.com>, "www-tag@w3.org" <www-tag@w3.org>, es-discuss <es-discuss@mozilla.org>
On Mon, Dec 2, 2013 at 6:24 PM, John Cowan <cowan@mercury.ccil.org> wrote: > Nico Williams scripsit: > >> "datetime" and such are interpretations of more basic datatypes, > > An interval of time is not a string, any more than a number is a string. > They are both *representable* by strings, but everything is: you can > represent a human being or the planet Saturn by a string. Interval definitely sounds like a tuple, but, sure, JSON texts can represent arrays as text... >> they belong in pre-agreed/documented schema rather than on the wire. > > The decision to do so is arbitrary. Of course it is. How much to describe on the wire vs. schema is... a continuum. Given the JSON we have though... it's best to deal with datetime via schema. My point was and still is that the one thing that's sorely missing is an indefinite-length unescaped binary data encoding, and which is not nearly as complex, notionally, IMO, as other types. Nico --
Received on Tuesday, 3 December 2013 02:23:38 UTC