Why I think ending article-processing charges will save open access