What is the shortest mRNA the ribosome can read to produce a peptide? This question came as a comment on a previous question regarding non-ribosomal peptide synthesis, and why Glutathione cannot be synthesized by the ribosome. In general, Glutathione has a "gamma" peptide bond, thus cannot be synthesized by the ribosome. However, is there an actual minimal length that ribosome can process? I understand that the mRNA is more then just the codons, however, does it still require a minimum set of condos, e.g., say 18 nucleotides, to produce a peptide?

Reading the paper linked by @canadianer and its references was pretty illuminating. It's a straightforward paper except for its conclusions, and the most convincing part is the evidence of actual translation. There's some evidence that transcription start sites are used as a regulatory mechanism by displacing other things that might otherwise bind. An unknown fraction, but most(90%+), very short bioinformatically detected ORFs are not translated.

Some, however, are. See this table for the experimentally verified translated micropeptides. The shortest is a nearly unbelievable 6 peptides long.