To understand why, we need to take a look at the history of biological classification.
The science of classification really started with Carl Linnaeus (1707-1778), a medical doctor from Sweden who is also known by the variant names Carl von Linné and Carolus Linnaeus. At the age of 28, he published the first edition of Systema Naturae, his ground-breaking classification of animals and plants - initially a single thin volume, although subsequent editions were to mushroom in size.
Before Linnaeus, classification was a haphazard business. For example, the common wild briar rose was referred to by different botanists as Rosa sylvestris inodora seu canina and as Rosa sylvestris alba cum rubore, folio glabro! There was little organisation of names: organisms were assigned to species, and species grouped into genera (for example, lions and tigers were recognised as separate species, but joined in a single genus, Panthera), but there were no higher level groupings.
As Systema Naturae progressed through its editions, it introduced two important innovations. The first of these was the so-called ``binomial system'', whereby each species is referred to by a two-word name, consisting of the generic name followed by the specific name - as in Panthera leo, or indeed Tyrannosaurus rex. This approach was in consistent use by the tenth edition of 1758, and remains more or less unchanged today.
The second and more far-reaching innovation was the introduction of higher-level groupings than the genus. In Linnaeus's original system, genera (such as Panthera and Canis, the great cats and the dogs) were joined in orders (such as Carnivora, the carnivorous mammals); orders (such as Carnivora and Chiroptera, the bats) were joined into classes (such as Mammalia, the mammals); and classes (such as Mammalia and Reptilia, the reptiles) into kingdoms (such as Animalia, which I shall not insult your intelligence by translating!) The key insight here was that life can be arranged into a branching tree-like hierarchy. The level-names order, class, etc. were known as ranks.
This simple five-level system (kingdom > class > order > genus > species) was soon augmented by intermediate levels such as the family, intermediate in level between the genus and order - for example, the family Felidae consists of all cats, whether great (Panthera) or small (Felis). As people found the need for more precision in locating how inclusive a proposed grouping was, along came superfamilies, subfamilies, tribes, superorders, infraorders, hyperfamilies, and ... well, the list goes on. Among these additional ranks is the phylum, intermediate in level between the kingdom and order. For example, the phylum Chordata, (animals with spinal a cord, including the vertebrates), is a part of the animal kingdom.
It doesn't take too much of this sort of thing before the system becomes, if not actually unworkable, then at least obscene. For example, in a Dinosaur Mailing List message, George Olshevsky has listed a hierarchy of thirteen ranks just between order and family!
For the more common of the lower ranks (superfamily, family, subfamily, etc.), taxon names have conventional endings - for example, family names end with ``-idae'' and subfamily names with ``-inae''. However, above this level, standardisation is at best partial, and varies across disciplines. For example, while bird orders all have an ``-iformes'' ending (e.g. the Galliformes), there is no such uniformity in the names of mammalian orders (which include Edentata, Pholidata, Lagomorpha, Rodentia and Macroscelidea among others.) In the same way, while bony fish have suborders ending in ``-oidei'', chondrichtyians (sharks and rays) have suborders ending in ``-oidea'' (which ending is conventionally used for superfamilies in the dinosaur world.) As if this isn't confusing enough, the rank names are used rather differently in different disciplines: for example, a ``division'' in plant taxonomy are broadly equivalent to a ``phylum'' in animal taxonomy, where ``division'' (when it's used at all) is more specific than class.
In the face of this maze of similar-sounding ranks, semi-standardised endings and contradictory ``rules'' (Quick! Which is more inclusive? A grandfamily or a hyperfamily?), one reaction has been a move to abandon named ranks altogether, simply naming nodes in the tree and making no judgement about how high or low level they are - and it is a matter of judgement rather than of fact; one man's parvorder is another man's nanorder. This minimal approach is particularly popular in the cladistics community, perhaps in part because the trees generated by cladistic methods have far too many nodes, and change far too often, to be amenable to labelling with ranks.
Proponents of this ``cladistic'' approach to taxonomy, then, would like to abolish all ranks above the level of species, which they argue is the only rank with an objective definition (but see ``When is a new dinosaur erected as a new species or genus?'' ). Under this scheme, even the rank of genus would be abolished. And therefore, of course, so would Linnaeus's binomial: every species would have to be given a new, unique name. Not surprisingly, these radical proposals have not met with unanimous enthusiasm - see ``Is the PhyloCode a good thing?'' for more discussion of pros and cons of this approach.
It remains to be seen how nomenclatural practices will change, but what seems to be emerging by default is an approach in which some of the more ``important'' nodes in the large, deep trees generated by cladistic methods get quietly labelled as families, orders, etc. according to a blend of historical precedent and what seems to be useful at the time. For example, most historical dinosaur families seem to have survived the transition into the cladistic age, and many new dinosaurs which do not fit into existing families have new families named after them - or at least, new groupings whose names end with ``-idae'', which is at least very suggestive!