In the latest post to the SitePoint PHP blog Bruno Skvorc proposes the idea that "legacy code is a cancer" that can influence decisions and technology choices that shy away from the new and possibly more functional alternatives.
This might come out controversial, but I firmly believe there is no room for legacy code in modern systems. Allow me to elaborate before you sharpen your pitchfork and light your torch. What I mean by that is: there should be absolutely zero reason to keep implementing the functions you're adding to the new version retroactively into the old version, just because some people are still using it, even if the people using it are a vast majority.
He talks about the "support everything for as long as we can" ideal and how it can come back to bite you in the end. He suggests that, at some point, the v1 users have to "be discarded" and dropped for the upgraded version of the application. He talks about failure potentially brining around success and compares applications versus libraries and components and the upgrade path for each. He ends the post with a suggested upgrade path to move the system itself away from legacy support and into the new, latest version.