That’s a reasonable idea. I can imagine a ribosome directly binding to DNA to translate the DNA codons directly into protein. We can’t really tell why that didn’t happen, as much of the early development of life occurred far in the past and left no evidence of the steps in the process; we must infer the process from observing modern biochemistry. However, I can speculate that much of the regulation of gene expression occurs at the RNA level, with variations in the stability of RNA, translational repression by mRNA, inhibition by natural antisense transcripts, etc. Perhaps the RNA intermediate was required in early life to provide a mechanism for regulation of gene expression.
What jonmoulton wrote is true (regulation on RNA level), but further there is amplification of the signal (there can be thousands copies of each mRNA, each giving rise to hundreds of protein molecules), also, during replication of DNA the protein synthesis would have to be shut down and there could be other reasons.
But as said, it’s always difficult to answer the Why-questions because there is no intelligence creating this stuff on purpose. It’s more like the blind watchmaker trying to repair the watch by random changes and "seeing" (as he’s blind) what works. It could be that in the RNA world the proteins were translated directly from the RNA (as nowadays) but when DNA started to work as information carrier, there would have to be too big changes in the ribosome and everything so it couldn’t be possible but rather the transcription system evolved (as that’s pretty similar to replication).