The best you can do is decode exactly twice and write off feeds where that’s not good enough as collateral damage. That will cover 98% or so for the feeds in the wild; the remaining 2% are simply unreclaimable, and whatever clever approach you employ as you try will just break a different (and larger) set of feeds.

The analogy to PPI is not very useful. Perl code has a precise meaning that can be read out unambiguously, even if it’s extremely hard to achieve that. In contrast, RSS titles do not have precise meaning to begin with.

As for Atom, I have no illusions about what it means for writing software that consumes feeds: nothing, because you clearly can’t ignore the millions of deployed RSS feeds. Nor was I saying you should at any point of this diatriabe. I’m just imploring those who are only just writing feed generation code now to please make the sane choice right off the bat, so that this headache, even though it is here to stay, will at least not grow (too much).

(Please note that it’s Atom, not ATOM: it’s neither an acronym nor a backronym and doesn’t stand for anything; it’s a proper name.)

Makeshifts last the longest.