* [EMAIL PROTECTED] <[EMAIL PROTECTED]> [2006-07-26 16:40]: > It has been suggested that I add a mutex to every SQLite > database connection.
Client code can easily use trivial wrappers to supply such serialisation for itself, though. > The downside is the (minimal) additional overhead of acquiring > and releasing a mutex on each API call. I don’t see this is a factor. A platform with threads is by definition not starved for CPU cycles. > There will also be some changes to the OS-layer backend which > will break private ports to unusual OSes. This is a big deal IMO. An idea worth pursuing might be not to implement such thread serialisation natively in SQLite itself, but to merely add a way for client code to supply the address of a callback to be called called at the appropriate time in each relevant function, so that API clients can implement serialisation on their own through whichever means they choose. If that doesn’t seem like a good idea and you’d prefer a deeper change that implements native thread serialisation, I’d suggest to merely earmark that for a time when a backend interface change is unavoidable anyway, and revisit it then. Regards, -- Aristotle Pagaltzis // <http://plasmasturm.org/>
