I think it's necessary to be pragmatic rather than purist about this.
Bear in mind that we always maintain bugfix releases in parallel with development releases; therefore saying we'll drop VC6 in the future means on a future development branch, not on bugfixes to the last release it was supported in. So including something that 'will be dropped almost immediately' is overstating it; the kind of people that don't upgrade their development environment will likely be quite happy to stay on 1.0.x releases.
Secondly; being realistic there are VC6 users out there who don't want to pay for an upgrade, and who will not want to switch to MingW. This could be for many reasons - such as the environment (IDE) being less friendly, that they are using other libs which don't work with MingW yet, and such things. You can't dismiss this, it will happen.
If VC8 (express), when fully released, is a truly free environment that can be used to compile Ogre and related libs, this argument may go away. But it doesn't yet - I have distinct suspicions that when VC8 is finished, the free version will be limited in some way greater than the beta is now. Dropping VC6 support before 1.0.0 could well be the deciding factor for these people to use something else. You might not agree with it, but it's their decision.
Lastly, supporting VC6 in 1.0.0 is not very hard, since we've already got there. In fact,
not supporting it is more work than supporting it at this stage, since to not support it would be to modify those parts that already have VC6-specific elements to them. Lets not overstate this either - VC6 support has not compromised the Ogre code in any meaningful way - it's just annoying for me (time-wise) to have to test it and put in new workarounds, so I'd like to drop it. But it's not hurting me enough to drop that support before 1.0.0, because that's the first release that we'll say is 'done', and for that, we should have a development environment that covers all the bases we have already.
I have no intention of stranding VC6 users on a 'beta' version unless there is a solid alternative - yes MingW is useful but it's not necessarily a replacement (as noted above), and the free version of VC8 is not finished yet and is essentially unproven, as far as I am concerned, both in terms of stability and final licensing terms.
I'm beginning to wish I hadn't started this thread - I was just interested to guage the current lay of the land ahead of time.