Re: [boost] JSON Archives for Boost.serialization

From: Jeremy Maitin-Shepard (jbms_at_[hidden])
Date: 2008-03-22 12:55:13

"Kasun Indrasiri" <kasun147_at_[hidden]> writes:

> Hi all,
> I'm interested in the Boost JSON archive for Boost.serialization project
> which is presented in the GSoC2008. I have had good exposure to C/C++ during
> my academic life and projects in the past.
> The main motivation to select the JSON archive project is that I have
> developed a JSON parser for Apache AXIS2/C web service engine and I
> successfully completed the project. In that case the parser was written in C
> but the task was similar to the requirements of the Boost.serialization.
> Therefore I would like to contribute to the Boost C++ community through my
> project.

It is important to realize that Boost Serialization, even when it uses
e.g. the XML archive format, is not designed to produce archives that
conform to any particular format/schema.  In particular, the only real
guarantee is that stream produced by serialization can be read back by
de-serialization if the same (or a compatible) version of Boost
Serialization is used and a compatible sequence of serialization
instructions is used.  In the case of the XML archive format, the file
produced by serialization happens to conform to XML syntax, but that
fact is nearly irrelevant as in all likelihood the archive will still be
easily readable neither by humans nor by any external tools (except for
trivial tools that merely display the XML structure).

Thus, it is not clear what advantage a JSON-format archive would offer.
The most obvious use for JSON is for communicating with a program
written in JavaScript, but then it would be necessary to follow a
particular format so that the JavaScript program could do something
useful with the data, and therefore Boost Serialization is not the right
tool for the job.

-- 
Jeremy Maitin-Shepard