> For Peter & Pepijn - I think the issue is essentially a forward-compatibility > problem moreso than a backward-compatibility one. So I think your idea on > introducing some version control would be the least painful. Indeed. The lack of rowid itself is not an issue. It's that someone could create a database using SQLite 3.8.x features, pass it on to someone with a SQLite 3.7 based app and run into runtime problems that they weren't expecting. We're looking for some way to reliably detect these kinds of cases or alternatively language to put in the spec to prevent this situation in the first place.
AFAICT there's no simple value in the database we can use to check for this. I mistakenly thought we could use the schema format number for this. Queries on sqlite_master could work but we would need to come up with a way to use this reliably. The suggested queries check for known problems. I'm worried that this will be an inherently brittle check since the next sqlite revision could introduce a new feature for which the spec has no checks in place. Perhaps we should make the allowed DDL subset a part of the spec. That way we make explicit what is allowed and anything outside of that is forbidden. Pepijn _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
