Transactional back ends

[scossu]

Hello,
As I dig into the persistence options available for RDFLib it seems like the only one that clearly offers transaction support is the SPARQL store. I am also very happy to see that the "dirty reads" option is available in the snapshot version, which is fundamental for me to perform atomic updates.

By looking at the code for the Sleepycat connector I am not clear about whether transactions are supported. If they are not, is there a reason for that (beyond that not being a priority)? If I were to extend the Sleepycat store class, or the Kyoto Cabinet one,  or even create a new one for a DB I prefer (e.g. I'd like to do that for LMDB) are there any caveats that I need to be aware of?

Thanks for your advice.
Stefano


EDIT:
By "supporting transactions" I mean being able to open, commit or rollback transactions from within my application after an arbitrary number of write commands, as I can do with the SPARQL store back end.