Though he often reveals previously not-obvious truisms, I have to differ with Robert Sayre on this, regarding support for Atom 0.3:
Standards compliance is not the goal. Uniform interpretation is the goal, and standards can help that along.
That picture still shows only part the story. The benefit of uniform interpretation is that it enables communication. But the generator of the material to be interpreted is also part of the system.
It may be possible to build say a feed aggregator that supports the 9 or so different varieties of RSS, plus their combinations and ill-formed variations, plus CDF plus MCF plus Atom plus Atom 0.3. Many aggregators approach this level of support. But consider the effort involved in supporting all those things compared to supporting just one. How many CDF, MCF, RSS 0.94 feeds are there?
If you follow through Robert (and Knuth's) argument, any piece of software will accumulate support for more and more things over time. This in itself is fine. But any new piece of software will have to support all those things too, which is costly. There have to be pragmatic considerations, and in that context deprecation is a Good Thing.
The goal is more like
good-value interoperoperability. Forget Atom 0.3, RSS 2.0
etc, forget even RSS 1.0 if you don't need the RDF features
directly. Provide good support for Atom and everyone wins.