>> Pretty much, yes. A full compatibility layer would like affect
>> performance big time and a less intrusive version is probably not worth
>> anybody's time to write.
>
>
> But D7 not will be a less intrusive version, this is the point.
>
"less intrusive version" was referring to the compatibility layer, not
Drupal.
Basically, we've had similar discussions for Drupal 4.7, 5, and 6 and
all were hailed as the end of Drupal as we know it because nobody would
be able to update their modules in time, etc yadayada. We probably have
had this discussions for even earlier versions but luckily I've managed
to forget them.
Executive summary: Compatibility layer == bad.
I agree, but it's bat too rewrite from scratch every time a lot of code, like Views or CCK or Panels.
Could be nicer have at least a stable API or a *drupalX-compat" plugin with older API, but bright people surely know this things better of me :)
So far, it has proven to be an acceptable price to pay for innovation. Yes, it
causes some pain, but if we fix the APIs between major releases, this
will be the start of the end for Drupal, since it will either get
bloated by a compatibility layer (more resources, slow, complexity,
...etc.), or Drupal will stagnate and no longer become a leading
content management platform/application.
Having said that, there is nothing stopping YOU (or others) from building a compatibility layers. Some people tried that from Drupal 4.6 to 4.7 after Form API was in core. We just don't want that to be in core.
Better focus on tools that would make migrating your modules easier/quicker (core, deadwood, documentation for converting from 6.x to 7.x, ...etc.)