I think one view is that a collection is homogeneous. And it appears that this is Mikes' view, correct me if I am wrong. In general, most of the documents in my collections could be described by a DTD. However, I have one collection that is a collection of different types of documents. I would prefer not to have to create an individual collection for a single document. I would also not want to be constrained in having only one type of document in a collection.
In one of my collections, I am storing RepairOrders. I also have a document that is a list of all the "Open" RepairOrders. I either have to have a very slick DTD to cover both or put my list in another collection. Seems artificial to have a single collection for one document. DTDs can be very useful when you are receiving documents from the outside world. They can help maintain the correctness of the data. They remind me of flexible schemas. I for one do not need any of the above or the overhead that comes with it. I don't have outside documents. I will rely on developers and QA for the correctness of my documents. And I choose XML because it gives me the flexibility to morph my document into any form that fits my Customer, not my DBAs or Developers. I see xml documents as data that comes in many formats. In so many formats that a DTD would be useless. +1 for NOT requiring DTDs and defining collections as being ANY document. DTDs as optional with no overhead for not using them is fine with me. Mike Mortensen wrote: > I believe that choice made (validating the collection) is the correct one. > It most closely represents what happens in the real world. Even with widely > different data and usage, it still makes sense to validate the collection. > > For example, if used in business, invoices generated and sent to customers > could be stored in Xindice. It is appropriate to validate the collection > since all documents are, by their nature, similar in content and structure. > The case is likewise true if Xindice were used instead to store the chapters > of a book. Each chapter has similar content and structure. We could just as > easily throw the periodic table of the elements into Xindice and science > would give "thumbs up" to the collection approach. > > I recognized that there may be occasions where the content is substantially > dissimilar. In this case, we simply put the documents into separate > collections and still get the desired validation outcome. > +1 for the path taken as the logical choice.
