If one studies a genome sequence and comes across a region that is of the length and arrangement of a protein-encoding gene, and is enclosed but not interrupted by start and stop codons, then one can reasonably infer that this sequence is likely to be functional, even if no other evidence is yet in hand for what its potential protein product may do. If someone doubted that this were actually a protein-coding gene, then additional evidence would be expected to show this, for example by showing that it is really an artifact, or a pseudogene, or indicating that it is probably not functional despite the characteristics that suggest that it is.
Many non-genic sequences that are conserved across taxa or which exhibit characteristics consistent with regulatory regions or binding sites or structural components may be treated in the same way as finding an open reading frame as noted above. For most non-coding DNA, there is no such evidence of probable function, however. In fact, as the Sandwalk series on “junk DNA” notes, most non-coding DNA is of a type for which there is little reason to expect function. Inactive vestiges of transposable elements and pseudogenes may occasionally be functional, but there is no reason to assume that they all are given what we know about how they form. Moreover, the massive differences in genome size (i.e., amount of non-genic DNA) among taxa, including species that would be expected to have similar regulatory, mutational buffering, or structural requirements, suggests that much of it is indeed lacking in a universally applied function.
The default assumption by those who accept non-adaptive evolutionary outcomes is that much or even most of a larger genome is not functional for the cell or organism level. This is because of what we do know about these sequences, not because of what we don’t know. Therefore the burden of providing evidence to the contrary is on those who argue that all or even a large percentage of non-coding DNA has a function, which requires an explanation for the variation in DNA amount among eukaryotes in addition to empirical evidence for function at least in general terms, if not as an indication of what the function probably is.