This chapter briefly lists features which are either not yet implemented or which don't work as they should or which work differently on different platforms. These things may be implemented or fixed in future releases, respectively.
The clients do not use a dynamic buffer to read the data from the application server. This may result in timeouts if you view the results with a pager that can't buffer large amounts of data in the background. To avoid these problems, it's prudent to redirect queries that return large amounts of data into a file.
The query results can only be sorted by ID (default) and publication year.
The length of the query string is limited to approx. 4kb. This can be increased at compile time.
There is no automatic duplicate check for references.
Transaction support is only available with PostgreSQL and SQLite, but not with MySQL. Therefore any action that changes the contents of some MySQL database may corrupt the database if something goes wrong. refdb of course tries hard that nothing goes wrong programmatically, but errors (and power failures) are possible.
Support for different character encodings depends on the database server. PostgreSQL offers a variety of different character encodings, including Unicode, which are selectable per database. MySQL currently supports only a limited range of character encodings which are set per server instance. Unicode will be supported in future versions of MySQL. SQLite offers Unicode support as a compile-time option.
Export of bibliography styles to BibTeX .bst files would be really cool but is not implemented yet.
The memory footprint of the application server could be reduced by moving code into a shared library.
There are no manpages yet.