Zero-impact classification plan

[Jon Bosak <Jon.Bosak@eng.sun.com>]

It's been difficult to get everyone onto the same call at the same
time, but I believe that the ad hoc subcommittee appointed to look
into the classification system for the reg/rep has agreed on the
following for the short term.

Jon

##################################################################

Basic assumptions

   1. The window of opportunity for establishing a vendor-neutral
      reg/rep for XML schemas and securing industry consensus
      behind this approach is narrow.  Therefore, we should do
      nothing that would prevent the prototype reg/rep
      implementation from launching on time.  Rather, we should
      aim to get something running as soon as possible and plan to
      continuously improve it in the months following its initial
      release.  In practice, this means (a) relying as far as we
      can on manual labor to populate the database and build the
      navigational interface on the portal site and (b) ensuring
      that the capabilities we will need over the medium term are
      in the database right from the beginning and then improving
      the UIs to this database in the time following the initial
      release.

   2. There is currently no standard classification scheme for XML
      schemas.  In fact, no standard classification scheme that we
      know of has as its objects anything even resembling XML
      schemas.  It is unrealistic to expect that we will be able
      to design such a system within the time frame for this
      project.  On the other hand, there are standard
      classification schemes whose objects resemble certain
      application aspects of XML schemas.  Three obvious examples
      of such systems are NAICS, whose objects of classification
      are industries; UNSPSC, whose objects of classification are
      products; and LCSH, whose objects of classification are
      publications.  XML schemas are often intended to apply to
      documents used in certain industries, or documents
      describing certain products, or documents whose subjects
      resemble the subjects of publications; so it is not
      unreasonable to expect that we can eventually use some
      combination of these axes of classification in order
      indirectly to classify and retrieve most XML schemas.

   3. Classification of submissions is a task for experts; it is
      not something that we can leave to the submitters.  What we
      need from the submitters is enough information for a
      classification expert (perhaps with the help of a phone call
      to the submitter) to properly classify schemas according to
      one or more of the systems we decide to use for this
      purpose.

Plan of action

   1. Ensure that the reg/rep database supports codes for multiple
      classification schemes and multiple codes within each
      scheme.  (The current implementation already does this.)

   2. Request submitters to enter relevant information regarding
      industries, products, and subjects of application in the
      submission's description field.

   3. Enable classification experts designated by OASIS to
      manually populate the database with industry, product, and
      subject classification data in the months immediately
      following the first release (and for test purposes before
      the first release).

   4. Improve the end-user UI in the months following the first
      release to make use of the classification data added by the
      experts.

We agree that we do *not* need UIs with classification scheme
menus in the first release.  These come later.

We believe that this strategy allows us to continuously improve
the classification of items in the reg/rep with no impact on its
launch date.

Jon Bosak
For the Ad Hoc Classification Subcommittee