How does C+ recombine in diffuse molecular gas?

Aims. We wish to understand the processes whereby the dominant state of free carbon shifts from C+ to C I and CO in progressively denser and/or darker diffuse and translucent clouds. Methods. We discuss recent compilations and observations of C I, H I, H-2 and CO measured in uv absorption and compare the observations with models of the thermal and ionization equilibrium including and excluding grain-assisted neutralization of atomic ions such as C+. Results. There are significant disparities in N(C I) and divergent behaviour with respect to H I and especially H-2 and CO in two recent discussions of the C I abundance in diffuse and translucent gas. If the older data tabulated by Wolfire et al. (2008, ApJ, 680, 384) are considered, the run of N(C I) with N(H I) and N(H-2) is comfortably explained only by models embodying grain-assisted atomic-ion neutralization, much as those authors noted. If the newer data of Burgh et al. (2010, ApJ, 708, 334) are considered, either lower density models with grain-assisted atomic-ion neutralization or much denser models without it may suffice. In either case N(CO) increases from 10(14) cm(-2) to 10(16) cm(-2) with little change in N(C I) in either dataset, presenting a real challenge to models of C+ recombination and CO formation in the C+ --> C I --> CO transition. Conclusions. N(CO) exceeds N(C I) even at N(CO) greater than or similar to 3 x 10(15) cm(-2), well within the regime of diffuse gas where the dominant form of free gas phase carbon is C+; one of the supposed signatures of the translucent regime, that C I is the dominant form of free carbon, is not found on the sky. However, the C I data clearly need to be put on a firmer basis before the C+ --> C I --> CO transition may be understood. Ambiguities in the C I column densities determined in uv absorption may perhaps be resolved by sub-mm observations with Herschel or ALMA.