- From: Tim Bray <tbray@textuality.com>
- Date: Sun, 10 Nov 2013 22:29:13 -0800
- To: Anne van Kesteren <annevk@annevk.nl>
- Cc: "www-tag.w3.org" <www-tag@w3.org>
- Message-ID: <CAHBU6ivwnrx+3mdXEEVU16wqSrhp_ASAA-SA9ce0wrYowCsW5g@mail.gmail.com>
On Sun, Nov 10, 2013 at 7:08 PM, Anne van Kesteren <annevk@annevk.nl> wrote: > To improve JSON interoperability the IETF should not define a more > restricted version of JSON than defined by Ecma-404. > The problem is that the IETF already *did* define such a restricted version, RFC4627, in 2006. So the interesting argument is whether, in 2013, any increased gain in interoperability due to removing the restriction overbalances the software pain of people who implemented “JSON Text” per 4627. My own feeling is probably influenced by my JSON experience, largely with usages in application-level network messaging protocols, usually RESTful in flavor. In this context the interoperability of JSON is superb, and basically all such protocols require that messages be objects (more commonly) or arrays (less so). So my inclination would be to retain things as they are. Also, the interop problems I’ve observed have all been around dupe keys, busted unicode, and programmers not understanding how computers store numbers [sad but true]. Also there’d probably be IETF community pushback against the -bis introducing an active incompatibility with the 4627, which the WG charter strongly discourages. But it wouldn’t be insane to raise the issue in IETF last call, which should be starting soon. > (Worth pondering about: what to do about a leading BOM, which > XMLHttpRequest and browsers allow, but neither IETF nor Ecma-404 > allow.) > Yeah. Maybe spirit-of-Postel’s-law language urging implementors of generators to avoid these things where possible, and implementors of parsers to tolerate them where possible? > > > -- > http://annevankesteren.nl/ > >
Received on Monday, 11 November 2013 06:29:42 UTC