So sayeth Mike Taylor (for my own confused ramblings through the quagmire of electronic publication, read my earlier posts on the subject). And this day presents us with a spectacular demonstration of that point.
In a paper in today's issue of Science, Ren et al. (2009) have presented an analysis of Jurassic to early Cretaceous long-proboscid scorpionflies and their role as probable pollinators of nectar-producing gymnosperms (as has also been suggested for kalligrammatid lacewings). As part of this study, Ren et al. present descriptions of six new species and two new genera of fossil scorpionflies. Nothing out of the ordinary here, except that (Science being Science, with its notorious restrictions on article length) the species descriptions are published in the Supporting Online Material.
From the point of view of the ICZN, Science is a perfectly valid forum for publication - thousands of copies are printed every week. But these printed editions don't include the online supplements, so the online-only component of the journal is currently not a valid publication. Technically speaking, the new species of Ren et al. (which are referred to and illustrated but not described in the print version) are nomina nuda. They are not valid names. But these online-only names have not appeared in some far-flung unfrequented corner of the internet, they have appeared in one of the world's most prominent science journals (like it says on the label). Their validity is going to be pretty much taken for granted.
Ren, D., C. C. Labandeira, J. A. Santiago-Blay, A. Rasnitsyn, C.-K. Shih, A. Bashkuev, M. A. V. Logan, C. L. Hotton & D. Dilcher. 2009. A probable pollination mode before angiosperms: Eurasian, long-proboscid scorpionflies. Science 326: 840-847.