Note that all archive classes currently included with the serialization library depend one std::streambuf to do the actual input/output and hence inherit all the attributes and limitions of the implementation of streambuf that you're using. SO...

1. is there any limitation on the file size generated by serialization. i.e. what is the maximum file size which can be written by boost serialization?

make a small test program which tests wheter std::fostream can create a file larger than the size you're interested in.

2. is there any limitation on the object size, which can be loaded/ unloaded by boost serialization?

currently - 1.34 - there is a limit of 2G objects in a collection. This limitation is being relaxed on 1.35 to handle up to (2G)**2 on 64 bit machines.

3. does boost serialization offer any way (checksum?) to ensure the file being written is not corrupted?

No. This COULD be done by deriving from one of the standard archives and adding on alittle bit of extra functionality. But, if this were my concern I would make a custom streambuf to do the job and use it with my archive.

thank you very much

You're very welcom Robert Ramey