Sohail Somani <sohail@taggedtype.net> writes:
On Sat, 22 Mar 2008 10:10:35 -0700, Jeff Garland wrote:
Programs written in C++ often need to inter communicate with programs written in other languages. Or they need to store data structures in a form that can be processed by programs written in another language. JSON is now a common format for doing this and has parsers any many languages.
The main problem as I see it anyway, is that even though there are many parsers, there is a boost-serialization-specific way to interpret the data. I think it is possible to write two different types of JSON archives: one that is meant to interface with the outside world and another that is just another proprietary serialization format.
If you look at the XML archive as an example, it is clear that any non boost-serialization processor needs to do specific things to understand the output. Specifically, the presence of object graphs is what I would see as the biggest hurdle.
I think if you want the JSON archive to interface with the outside world, you should forgo object graph support. Or atleast support both modes.
I think really Boost serialization just isn't the tool for the job if you want to produce an archive that can be read by something other than boost serialization. -- Jeremy Maitin-Shepard