Actually, in the gene-tech community it has been acknowleged for some years now that non-coding ('junk' to the layman) DNA does indeed serve a purpose in higher order organisms. Parts of this material really is junk insofar as we can tell, and all of it is in many plants and simpler organisms. However, even though it may not code for a protein, it often still serves a function - regions of it act as 'error magnets', as certain replication errors can travel along a DNA strand and due to their exact electronic configurations certain sequences of bases can 'trap' such errors and prevent them affecting critical gene coding regions. Furthermore, short portions of these segments code for short fragments of RNA whose sequence is a match for a short part of an mRNA strand and can bind to it, thus disabling it. These fragments, called iRNA (I believe, I'm out of touch now) 'interfere' with the process of reading RNA by the ribosomes by literally blocking part of the RNA. By activating the production of these pieces of RNA it becomes possible to regulate the expression of the mRNA, giving fine-tune control over gene expression. This tends to be observed only in more complex organisms, and the effect is (I would guess) a happy coincidence arising from the accumulation of a bulk of useless non-coding DNA, which has since been tuned by evolution as once it emerged it would modify the organism's structure just like any gene.

A more interesting thing to note is that what was once held as a large quantity of irretrievably corrupted genetic data on the Y chromasome was in fact a mirror of the coding genes which are Y-exclusive (the Y chromasome has some genes in common with X, as well as some unique ones. Sorry ladies, it ain't just a cripple X - guys actually have greater total genetic complexity, at the expense of some redundancy). Recent research has shown that the Y chromasome can fold back on itself in a curious variation on standard genetic auditing processes to self-correct on the grounds that the 'non-coding' region, being accessed less often, is more stable. Which bit is the non coding bit is readable due to the differing orientation of the deoxyribose 'spinal' units of the helix.

Sorry for the Xbox post.