Re: [xml-dev] Is it time for the binary XML permathread to start up aga

[noah_mendelsohn@us.ibm.com]

Ed Day writes:

> There is JSON as well.  I'm not sure what format it uses under the 
> covers to encode the data though.

JSON is good at solving the particular problem of sending pairs of 
named/typed fields and their values, where the values can themselves 
(recursively) have that same structure.

XML is aimed at a much broader class of uses.  For example, while one can 
niggle about the details,  XHTML does a pretty good job of conveying HTML 
in the form of XML, including all the mixed content stuff like <p>My point 
is that this paragraph has <emph>mixed</emph> content, in which markup 
occurs within strings.</p>  JSON doesn't even try to do that in a standard 
way. JSON also doesn't do a lot to support the distributed invention of 
cosmically-unique names, as namespaces do.

Therefore, JSON is in some important ways a step toward convenience and 
simplicity for the many cases where what you want to do is easily ship 
around a bit of data, and where you don't need a lot of distributed 
maintenance of the names of the fields.  XML is more complicated, but as 
applicable to a much broader range of uses.  I'd be very surprised if JSON 
were a suitable encoding for XML in general.  For example, I don't look 
forward to seeing XHTML rendered in JSON.  If an efficient XML encoding of 
any sort, including the ones Roger lists, purports to carry Infosets in 
general, then it should at least do a good job with compound languages 
built of XHMTL, SVG, etc.

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------