Whilst recognizing that it is undesirable (and impossible) to exert control over 'who' might provide or create taxon concept definitions, it would clearly be beneficial to avoid an explosion of taxon concepts by discouraging excessive representation of 'poorly defined' concepts and the multiple representation of 'the same' concept in various slightly different formats by different editors.
In some respects the definition of concepts should be self regulating: the majority of definitions will come from available electronic taxonomic datasources and database providers who have an interest in accurate and non-redundant representation of their concepts, furthermore the work required to create or record concepts de novo will not be trivial.
Once concept definitions are available and can be represented by resolution applications through the TCS format (and possibly even referenced by GUID) biologists (data creators) will be able to use these concepts to mark-up their data (by GUID or key references). However, if no suitable GUIDs or concepts were available to a data creator who wishes to reference taxonomic concepts in their dataset, they would be forced to define their own concepts as part of the data record. Where created by non-expert taxonomists such concept definitions may be quite poor or incomplete records of concepts. The originator of the data may only be able to provide limited information about the concepts that they are using (for example species name + identification source). This might not represent a true Taxon Concept AccordingTo a 'recognized' punblication or source. And it might not be desirable for such 'poor' or incomplete concept definitions to form the basis of future work, without some form of taxonomic verification, or record completion.
In these circumstances it might be useful for the concept record creator to be able to flag the concept as 'invalid' or 'provisional' etc, and only for taxon concept records from 'bona fide' taxonomic resources to be flagged as 'valid'.
The extent that this would require self regulation or a more formal registration could be determined by bodies who issue GUIDs for concepts.
It is clearly stated in TCSAndTheLinnaeanCore that
The TCS schema should be able to represent information in the LinnaeanCore model of names, primarily by expressing relationships between concept (which have names, and therefore can capture relationships purrely between the names of concepts).
Still I have difficulty in understanding of this model. Suppose a author defined a single concept with multiple names in a publication. It can happen in description of the concept in a language with both scientific name and the concept specific name in the language. There is a single concept, and multiple names (i.e. multiple relationshps between the concept and each name literal). If we prefer to have single GUID to each concept, why do we design to assign multiple GUID to single concpet only for multiple names? If the concept can be defined, or at least identifiable (GUID of a concept assumes that the concept can be idnetified), why we can't use NameUsage as a wrapper of TaxonConcept? In the above case, each of multiple NameUsage contains reference to TaxonConcpet, but unnecessary TaxonConcept itself. Of course this TaxonConcept may have reference to the original NameUsage if known. As summary, I suggest to add NameUsage (or something equivalent whatever its name literal) as a root element of the schema, transfer name-relevant elements in TaxonConcept v0.72 to NameUsage, and give a ReferenceType element to the NameUsage to refer to TaxonConcept. This design doubles necessary GUIDs at least, but enable to keep TaxonConcepts GUID space smaller. Note that this design based on distinction between name and name literal; name is treated as a relationshp between a name literal and the target object, rather than name literal itself. We have difficulty in names without this distinction. -- JamesYtow, 01-Sep-2004
Clearly you model concepts one way, as name usage. Our abstract model can capture your name usages concepts, that is why it is inclusive, i.e it can handle your way of representing concepts. It is not designed to exactly fit or prescribe one representation of concepts. So you can continue to represent name usages as you do - but could now communicate your data with people who model concepts differently, through the TCS schema. -- TrevorPaterson 2 September
No, TaxonConcept is a subset of NameUsage (see 
http://www.nomencurator.org/TDWG.html). The latter also willingly covers excessive representation of 'poorly defined' concepts and the multiple representations of 'the same' concept discouraged above, if they are different NameUsages. NameUsages are record of facts that the name is used, while TCS sounds more restrictive (although relaxed in some areas as AccordingTo is not required element). Multiple reprensetations of 'the same' NameUsage can be detected as multiple records having the same combination of name literal and AccordintTo. On the other hand, detection of multiple representation of 'the same' concept requires clear definition of the concept. What we can record in database is only NameUsages. TaxonConcpets are humans' interpretation of these usages. NameUsage may have data elemnts relevant to definition as relationships with other NameUsages or Circumscriptions sensu TCS. Without such 'definition' the data still name usage but unnecesary to be concept; so why do we need to call it as concept? What we need is to distinguish name literal from name (or name usage, to avoid confusion with multiple meaning of the word name). Usage of a name (literal) in a context (e.g. publication) must carry a concpet (otherwise it doesn't work as a name), but it is unnecessary that the concept recorded with its concept definition. I prefer to encouraging richer NameUsages rather than discouraging poor TaxonConcpets. Partially sociology? Could be, but not only sociology. -- JamesYtow 02-Sep.-2004
We suggest that Taxon Concepts would only represent published taxonomic opinions (i.e. where someone does a taxonomic consideration of the taxon and publishes in a taxonomic resource - "in my opinion this taxon A is defined as x,y, z."- eg published revisions - not every potential taxon). What people consider are 'taxonomic resources' are up to their value judgements - but are pretty clear to the taxonomic community. e.g. they would probably not be field guides but would be the monographs etc that the field guides were based on, and they definitely wouldn't be 'datasets' recording observations about a concept. -- TrevorPaterson 03 September
It is my understanding of TaxonConcept in the TCS. However, one of advantages of the TCS is allowing variable atomisation e.g. a TC can be specified by NameSimple only. It enables the TCS to represent published things other than taxonomic opinions. Dinstinction of a taxonomic opinion from its usage is hard task for other than taxnomists, and taxonomists are busy to refine their concepts where such distinction is required. What we need in context of GBIF is a schema enabling data exchage beteween data resources even covering taxa where valuable judgements hard to obtain. I strongly suggest to change the name of data element to NameUsage. It doesn't harm the shcema at all. -- JamesYtow 03 Sep. 2004
I'm sure this can be discussed in NZ - and if enough people want a change.......... --TrevorPaterson