[Insight-developers] Removing method deprecates as of ITK, 3.6

[Simon Warfield]

>
> I'm still uncomfortable with removing a method once we have released
> it. I agree that we should certainly warn if the method is used. But,
> if it does not pose a significant maintanence problem, why should we
> remove it? Once it's removed, there is no way to notify the user of
> the method other than a compilation error. These errors are typically
> cryptic.

  In a young toolkit, with many good features, and some bugs and API 
mistakes, I think it is good to learn how to do it right from experience,
and then to remove the cruft so as to make it easy for developers to 
understand what are the right choices and what are the wrong choices.

Old code that was deprecated should be out at major version updates.  
The older releases of the software don't stop working just because the 
new version is not held back by earlier decisions, so users adversely 
impacted by the evolution of the toolkit can upgrade at their own pace.

Here are some entertaining articles that address the same issue in other 
code bases:
http://www.joelonsoftware.com/items/2008/03/17.html
http://buytaert.net/backward-compatibility

-- 
Simon