Thanks for all your answers. I'm grouping the response here. Regarding SAVEPOINT, I think that if they are not supported at all, they should be disabled! Maybe add a flag to give access to all the unstable/uncomplete features, but disable them by default. That could be either a global setting or a per-connection setting. This way, no risk to use a feature that may look like it is supported. You said that it was documented that the feature was not supported, but from what I understand there are two SQL:2003 features involved: - T271 (Savepoints) - T272 (Enhanced savepoint management) But in this page: https://www.monetdb.org/Documentation/Manuals/SQLreference/Features/unsuppor... only T272 is listed as not supported. Should this imply that T271 is supported? Isn't that what we were using in our bug report? Regarding our internals patches, we discussed that a bit when we meet some of you in Paris. We plan to release our patches if they can be useful to the community. We're a shop that make heavy use of Open Source, and in turn, we're releasing various components to the community (including our core business engine.) Overall, we're pretty happy with reactivity when we fill a bug: it is fixed in few days! And that's awesome. Part on our interest for MonetDB come from that. The projet is well alive and actively maintained. Nonetheless, the nature of the bugs, which seems to touch "basic" stuff each time (query that are not uncommon) make us doubt about the targeted audience of MonetDB. Sure, many of you are using MonetDB in production. My article title was a bit harsh. But since we're building an application around it, and that we want a certain assurance regarding the result, we grow some concerns. Especially because we're building the query on the fly from a abstract, dynamic, definition, thus we can't anticipate all the sort of combination we will get (UNION, GROUP BY, aggregates, lot of subselect, ..) and thus might trigger a silent bug returning incorrect data for example. But don't get me wrong: we love the MonetDB for its design, its speed and its ability to be extended. Regards, -- Frédéric Jolliton Sécuractive