On May 16, 2005, at 10:11 AM, Robert Ramey wrote:
This raises a question that comes up from time to time. Where should code for serialization of other types be placed? An who should be responible for it?
I presume that serialization code should always reside with the data type it serializes, unless it is impossible for some reason (e.g., cannot modify standard containers)? As for responsibility... whoever is interested will end up being responsible, since we're all volunteers anyway :)
I would hope that individual library authors would take on the task of implementing serialization for the data types that they support. This has worked well with data-time, multi-index, and others. It hasn't worked for variant or shared_ptr.
It'll happen, I'm sure. I know I'd like to do it for the graph library, but it's a rather large undertaking and I haven't had the time. I'm sure other library authors and maintainers are in the same situation. Doug