It's not there. boost::archive works sequentially on a stream and that's it, so if you need to skip large portions of data you will have to divide your data into multiple files and/or multiple chunks in one stream. I did the latter and it works fine if you build some kind of meta-information structure on where chunks begin and end. This enables you to seek a specific chunk of data and process that with an archive.

In any case, if you want to improve performance of your app use (portable) binary instead of xml archives!

You could take a look at http://code.google.com/p/hiberlite It will do what you need storing objects in sqlite database. There were no new releases for a couple of months, but I'll return to this project in a week. And there are _already_ enough features to solve the "selective access" problem. Yours sincerely, Pavlo Korzhyk ICQ#155870780