Re:Re: [boost] Serializing bitsets and tuples

From: Robert Ramey (ramey_at_[hidden])
Date: 2004-09-16 13:30:20

>> I'll see if I can get a test case thrown together for std::bitset.
>> I'll take a look at boost::tuple as well (but that should be big time
>>  trivial, so... :) )

>Erh... trivial, but tedious. Seeing as I have to redo the serialize
>funtion for all possible lengths of a tuple (0 to 10 contained
>elements). Unless I missed something obvious? (again)

>Anyway, is serialization for tuples supposed to be a Boost.Serialization
>add-on or a Boost.Tuples add-on? As Jeff Garland mentioned, generally
>std::* should be in Serialization, and anything else in the respective
>library. However, Boost.Serialization already provides support for e.g.
>shared_ptr<> and Boost.Optional. So...?

This question has come up before without being explicitly/publicly agreed
upon.  Jeff Garland and I have come to the following consensus:

Serialization for standard containers and objects, ie members of namespace
std:: will go into the boost::serialization namespace and their
corresponding test will go into the serialization test library. 

Other serialization implementations will go into the namespace/test set of
their respective libraries.

All maintainence of serialization modules will be responsibility of their
respective authors.

There are a few boost classes in the serialization library which ended up
there during development of the library. shared_ptr is and example.  These
are anomalous cases and not a pattern that will be continued.

Obviously, its my desire to keep the serialization library as task of
O(1000) rather than O(1000 + N).

So that's our policy.

Robert Ramey