The Extended Monotropic Programming problem deals with a more general extension to all n variables at a time than the two variables at a time extension you are interested in, except that the literature I have seen deals only with the constraint set $C$ being a closed linear subspace, not a general convex set. Specifically: $\min_{x \in \text{closed linear subsapce}} \Sigma_{i=1}^nf_i(x_i)$ with respect to $x$, where $x = (x_1,...,x_n)$, $x_i \in \mathbb{R}^{i}$ See ["Extended Monotropic Programming and Duality", Dimitri P. Bertsekas, Journal of Optimization Theory and Applications volume 139, pages209–225(2008)][1] 2010 postprint available at https://www.mit.edu/~dimitrib/Extended_Mono.pdf Additional reference: [On a zero duality gap result in extended monotropic programming,Radu Ioan Bot, Ern ̈o Robert Csetnek"][2] Older published version: https://link.springer.com/article/10.1007/s10957-010-9733-y [1]: https://link.springer.com/article/10.1007/s10957-008-9393-3 [2]: https://arxiv.org/pdf/1002.3272.pdf