I am playing with the idea of creating a libodf library so programs in Linux can preprocess ODF and dump it on standard output.

 Extending this as an input preprocessor for system commands like less or future odf implementations.

 The libodf could be based on libxml2 and libzip to handle packaging. However there are common challenges on every odf implementation as to how to handle buffering and speed improvements.

 Also how to treat the namespaces, if we should just focus on active namespaces or the whole ODF collection.