Chatham wrote:

But certainly the choices made in the legacy documents
could not have foreseen emerging information standards. Even for
contemporary documents, the information standards that we use evolve.
Anyhow, there are still a huge number of legacy documents out there, and
even the best education on standards to enhance ongoing descriptive
practices and tagging practices won't change that.

I wholeheartedly agree with what Chatham is getting at here, and I think "education" is a key word here. Not to make a broad generalization about the community as a whole, but in my discussions with other archivists interested in implementing EAD, their perceived importance of the data content standards seems minimal.  The key is simply to put existing finding aids - - with as little effort as possible (i.e., little or no re-engineering) - - into EAD, irrespective of the content standards.  Thus, we discover a key cause of "tag abuse".

In regard to the tone of many of the recent postings, I agree that the flexibility of EAD is a benefit.  It allows many different types of descriptions.  As Michael Fox pointed out, it is also beneficial because it plays nicely with international descriptive formats.  

Just to clarify my stance on the issue, and one that our archives has taken, I think the use of EAD is perfectly acceptable for the creation of *new* finding aids (i.e., descriptions that adhere to existing data content standards created in light of its future application in EAD).  The only structuring I'd like to see is some kind of formal application and minimal encoding guidelines, whereby a standardized template is developed and put into practice.  The obvious benefits of this are that software can be created to enable the creation of EAD instances, and that users may be able to predict the presentational structure of information as they consume it. Thankfully, as the other list members have stated, several institutions are in the process of developing new guidelines (and making significant improvements to existing ones).  

The flexibility of the DTD, however, also creates time (and cost!) consuming confusion as to how to apply the loose structure of the DTD onto a given body of finding aids.  This is only exacerbated when trying to apply EAD to a wide array of descriptive styles within an institution or a consortium.  The enabling capabilities of the DTD, coupled with the exact points that Chatham makes regarding the promulgation of standards and existing education, has led me to question EAD's value for encoding *legacy* finding aids.  Chatham correctly points out that the creators of legacy finding aids could never predict the future paths that descriptive standards would take.  Nor can we really predict what descriptive standards will contain in 15 years.

Given the fact that a fair number of institutions already possess HTML (or some format of word processed documents that can be easily converted into HTML), I wonder if we might be better off recommending that their documents simply adhere to some form of XML.  Specifically, the use of XHTML would be straightforward here, as tools such as Tidy exist to clean HTML validation errors to a certain extent.  The beauty of using XHTML is that we can still use RDF and/or Dublin Core namespaces to provide further intellectual description (separate from the presentation) where necessary.  Similarly, if I understand the scope of the project well enough, it seems that the Research Support Libraries Programme Collection Description project (RSLP CD) could also be applied in lieu of EAD.

This is an approach we're considering here at AIP.  Although we'll never know if the descriptions we create here at AIP will adhere to standards 5, 10, or 15 years from now, I think it's more prudent to not reinvent the wheel for each finding aid that we already possess in some digital format.  I can clean the document in less time to adhere to some XML DTD or Schema than I can to encode it all over again in another format.  By doing this, maybe I can learn a little bit more about making standards-compliant finding aids, which should be a main focus of any EAD project.

Maybe a discussion of "EAD Lite" should entail coming up with a smaller set of repeatable DC elements or attributes that can be repeated in XHTML for the conversion of legacy finding aids.

Just my wacky $.02, I don't know if anyone else has thoughts like this!