Consider removing the dependency on Qt altogether for this part by using Boost directly, which is more stable and can be ported more easily, reducing the dependency tree.

Dropping QT from libkolab would be a complete rewrite of it, and also mean we no longer use any KDE libraries.

The most used facility of libkolab is the implementation of the MIME message format, which could be implemented using another 3rd party MIME library.
Other parts such as the calendaring facilities would require an alternative to KCalCore.

Overall this would be a significant change in our strategy. In case Qt is indeed too much to bear, then we should probably rather look into writing another library for the parts that we require that has no ties to qt, like libkolabxml.

Or did I misunderstand the suggestion?