JSON parser grammar
david-sarah at jacaranda.org
Mon Jun 8 18:56:13 PDT 2009
Douglas Crockford wrote:
> David-Sarah Hopwood wrote:
>> U+06DD and U+200B (format-control), and U+FDD0..FDEF (noncharacters)
>> should be added to the list. The rationale for adding them is the same as for
>> the other format-control characters and noncharacters.
> The rationale for the set the json2 escapes is that those characters are
> mishandled by one or more browsers. That rationale does not hold for
> your set.
ES3 specified that all format-control characters should be stripped.
Therefore, format-control characters cannot interoperably be passed to
'eval'. If some browsers failed to strip U+06DD and U+200B, then that
was a bug.
Noncharacters similarly cannot interoperably be passed to 'eval' because
th Unicode standard specifies that they shall not be treated as characters.
It is therefore noncompliant to Unicode to accept them in source code.
(Arguably, this means that they should be explicitly prohibited in ES5.)
As John Cowan pointed out, it is quite harmless for the set of escaped
code units to be conservative; none of these are commonly used characters.
David-Sarah Hopwood ⚥ http://davidsarah.livejournal.com
More information about the es5-discuss