Well there's also another point to separating the data layer. Separating data all together allows to use any data source, not just a database. For example, if the function calls and results were uniform across the entire data layer, it could allow usage of flat files or XML/RPC calls, not just a database. If a particular data "driver" choses to use a database, it can be any libary, including MDB2 or PEAR:DB or direct calls to that database.
On 10/20/05, Jeremy Jongsma jeremy@jongsma.org wrote:
Another consideration.... RoundCube is in *alpha*. MDB2 will likely be fully released and stable by the time RoundCube is.
-j
On Thu, 2005-10-20 at 15:39 +0200, Jeff Nichols wrote:
I don't really have many useful opinions to contribute to the database abstraction discussion, but I do personally disagree with the following statement:
Script Head wrote:
One thing that should be considered befor switching to MDB2 is that
it's
sill in beta. If RC was to reach a stable version, using a beta
database
layer isn't the way to go.
Using a beta framework is not necessarily a bad thing for a stable product. Think of it like this:
Say 'Hypothetical DB Abstraction Framework' (HDBAF) is really really great, but they're still in Beta because they're trying to work out how to support 'Mainframe Huge ObscuOld Database'. Well, we don't really care about that database, so there's really no reason not to use HDBAF as it is. *As long as you test it for the functionality that you need.*
Of course, you do have to worry about incompatible updates to the framework. But that's true of stable frameworks as well (it's just a word after all), and if you include the framework with your distribution, you can include any version you want.
Just my outtake on beta frameworks and libraries.
Jeff
-- Jeremy Jongsma jeremy@jongsma.org http://jeremy.jongsma.org