A good introduction into this topic can be found at wikipedia:
http://en.wikipedia.org/wiki/Junk_DNAAs explained there, the term "junk DNA" is now considered to be not accurate. What is missing from that article is a list of different kinds of junk DNA, so I'll try to make something up here. It probably will not be complete, as I am just going to rely on my memory at this point:
1. Non-coding parts of gene. These are regulatory sequences. In essense, these are "cond" part of a gene. This DNA is very well represented in DBs.
2. DNA maintenance sequences. Centromeres and telomeres - these are needed to keep chromosomes from damage, to regulate their separation during division and other maintenance functions. There is no need for these sequences in DBs, cause maintenance is automatic.
3. Repeatative sequences. These are essentially viruses that got messed up, so that instead of killing the cell and coming out, they got stuck inside the cell. They kept "re-infecting" the same cell over and over, inserting their DNA into DNA of the cell in many positions. Many of these are non-functional. DBs does not have this, but we have something similar with viruses - they can accumulate with time. These sequences don't add much to evolution, they are really just parasites.
4. Pseudogenes. These are genes that got messed up in some way and they just sit there, slowly (sic!) accumulating mutations. They can come back to life, most likely through a recombination with an active copy of the gene. The analog in DBs is a gene that has a messed-up condition, like "1 2 =". The problem is that due to high mutation rate in DBs, these genes accumulate mutations much more rapidly than they do in real organisms. Also, we don't have a workable recombination mechanism. In this category would also go remnants of genes - not complete genes, but pieces large enough to code for something. We don't have this kind in DBs and it would be nice to get it.
5. True junk. Seqences that either never coded anything useful or mutated so much that there is no useful information left at all. it is unreasonable to expect any useful thing to come out of these sequences, because they are absolutely random. Even if you start to express them, you would not make a protein. This is the kind of stuff that is analogous to "monkeys and War and Peace" idea. DBs does not have this and I don't see why we would need this. In fact, we may be better to actively avoid this.
Maybe I forgot something, but these are the major classes of junk DNA.