From: Scott Woods (scottw_at_[hidden])
Date: 2003-05-27 21:56:23
> It isn't clear from you post if you are aware of all the work (including a
> formal review) that has already gone into persistence and/or serialization
> at Boost. See the archives at
Sheesh. A lot of work has been done. A comparative analysis would
take some time. Also, if an answer exists then it might be of the form
"there is no silver bullet, rather technique Xyz will suit...".
If anyone else is interested in (or have heard about) a variant-based
persistence mechanism then please drop a message to this list. That
seems to be the only distinguishing thing about the approach I have
It has resulted in some nice properties, e.g. loading from a file
into a variant of undefined type sets the type. The variant can then
be tested for what object type was found.
It also handles unknown types (hashed id not known to loading
application). An "opaque" type is fabricated and supplied to
the loading application. Applications can trade data files (or
send network messages) in a type-safe and robust way, i.e.
transmission of a type the receiver doesnt know about is non-
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk