For a long while I felt oddly bitter that Apache AxKit, a pluggable XML transformation pipeline and web application framework, never enjoyed much attention. I always wanted to use it for serious work but could never justify it owing to lack of mindshare. I was never an XML advocate, but I never hated it either. Ditto for XSLT. AxKit just seemed so elegant and
practical, smoothing over some of the problems with XML and XSLT and complementing them in a way that made their promise seem more attainable.
For a startup I vetoed Ruby on Rails in preference to an AxKit-like approach, extending the PoC I had originally cobbled together. In retrospect it was incredibly stupid. Most importantly, I should've just let our dedicated web developer use what he was comfortable with, even though I had strong opinions about the usefulness of that approach. (Object binding had like near zero utility in our particular case.)
Like with ColdFusion is was a hard-earned lesson in understanding that the best tools are the ones that maximize the productivity of your actual and prospective staff, not some hypothetical 10x coder. It's a seemingly obvious principle, but not so obvious and easy in application. It turned out that Ruby on Rails would still be on the upswing for some time afterward, so vetoing it was especially dumb. But ColdFusion, which made sense in context for similar reasons, was doomed to flame out quickly and so even if I teleported my post-RoR-veto self to my first ColdFusion job, I would have made the wrong decision as I had actually successfully advocated moving away from ColdFusion, the accidentally correct decision.