Bar-coding is a quickie, simplistic answer to a very difficult problem.

I do not approve of barcoding as it is currently used. It is generally based on a much too small sequence (often part of the mtDNA COI gene) which is unlikely to be a fair representation of evolutionary history, which it is meant to model. Furthermore, as a statistician as well as an evolutionary biologist (molecular phylogenetics) I find the cut off points for delineating species to be simply arbitrary and based neither on any statistical or biological underpinnings.

Defining species has always been difficult simply because the evolutionary process is a continuous one with no well defined boundaries when populations suddenly become species. And a little bit of DNA is unlikely to provide enough information.

I think that the best that can be done requires much more work including understanding the ecological, morphological, and demographic information along with a solid tree built from a variety of well-chosen nuclear and mt DNA sequences of some length. One might consider that clades of reasonable branch length and support to be a good guideline for defining species. But even that is fraught with the problem of different rates of evolution among different organisms and stochastic error regardless of the method.

As George Gaylord Simpson pointed out many years ago, delineating species is an art and requires solid science combined with biological understanding and the results are never more than what a certain scientist believes are separate species.

While the ITS gene region is generally regarded as the barcoding gene of choice for fungi, a combination of gene regions such as ITS, Tef and beta tubulin would enable a more accurate taxa determintion and delineation.