[Dbix-class] Does anybody actually care about DBIx::Class::CDBICompat?

Darren Duncan darren at DarrenDuncan.net
Sun Nov 20 06:27:34 CET 2005


At 3:25 AM +0000 11/20/05, Matt S Trout wrote:
>Seriously, though - I've heard of very few people actually using it, and
>several suggesting dropping it.
>
>Does anybody use this? Was anybody planning to try and use it as a half-way
>house for porting?
>
>Even if it only ever has a few users, if people are relying on this in their
>current or planned code I'll happily keep it around - but if not I figure
>I'd be better spending my time elsewhere ...

My suggestion is this, assuming that CDBICompat already works right now:

0. Make one more combined release, asap, that has both the 
DBIx::Class native features and the CDBICompat add on layer being 
fully functional or at least as functional as they ever were, in the 
vein that you have done to date.

1. Split the current combined distribution into two, with CDBICompat 
going its own way, such that the combination of the two is as 
unchanged as possible besides the fact they are in 2 separate 
distros; a more or less pristine test suite should pass as it did 
before; if your combined distro was well organized before, this may 
mean little more than arranging existing files into two folder trees 
and customizing the Makefiles.

2. Following that initial and fully functional split, go ahead and 
make whatever changes to DBIx::Class' native API and functionality 
that you wish, without regard for whether the initial separate 
CDBICompat release would be broken by it or not; at this point, 
anyone who wants to can maintain CDBICompat separately and keep up 
with the core changes, and you don't have to worry about that.

2b. This said, assuming that DBIx::Class is meant to be a successor 
or replacement for Class::DBI, you will want to retain enough 
customizability in your native API that it is actually possible to 
continue emulating Class::DBI on top of it when someone makes 
appropriate CDBICompat changes.

2c. Rather than being core functionality, you *may* find it useful 
for CDBICompat to be maintained since that would be an extended test 
suite of sorts for your native API.

-- Darren Duncan



More information about the Dbix-class mailing list