Tim Bunce wrote: > > On Fri, Mar 22, 2002 at 02:59:00PM -0800, Jeff Zucker wrote: > > > > My previous thinking was more along the lines of providing only the > > basics so that no one would be tempted to use it when they really should > > be switching to DBI. > > I think the performance drop will take care of that :)
What about those pure perl drivers that include their own database/file management engine -- CSV, AnyData, Excel, XBase, Sprite, Chart? (are there others?). With these DBDs, the basic limit is the engine's searching through the files and that will be the bulk of the slowdown regardless of how they're accessed from the top. Does DBI itself (as distinct from DBI + drivers for real dbms's) really provide enough magic to provide a significant speedup for these low-end drivers? (a change from worse to less worse perhaps?) I do understand that providing a full DBI API for all pure perl drivers in pure perl (DBI::PurePerl) is much more elegant and far reaching, but I'm not sure that project would have the same goals or criterion as one with the more limited aims I have been thinking about (DBI::Lite) -- providing the subset of DBI and its drivers that is guaranteed to work on all platforms without compilation, with older perls (and hopefully newer perls), and without requiring other modules outside of core perl or any other external programs. I think Simon Oliver best expressed the goal: "the perl community would be guaranteed a working relational database on all platforms, much like SDBM guarantees a working DBM." And even though SDBM is a toy compared to other approaches, its ubiquity is a feature in itself. -- Jeff
