A colleague suggests that the secret agenda is to complicate MARC to the point that it collapses.Only that MARC is incollapsible. The software base that understands and handles MARC and nothing beside is just too large, to replace it and to migrate into new environments globally is an intractable task logistically and financially. More and more systems will of course be left behind and not be able to make full use of all new or changed fields and features.
Something new can emerge and grow only alongside MARC, which will then slowly (but very slowly) fade away into oblivion some time next century. For a considerable period, there will thus be an uneasy divide in librarydom, and this will leave the bibliographic universe in disarray for even longer.
Interesting comments from both. It seems to me that MARC could be much more flexible than it is now (returning to my rant on getting rid of ISO2709), but the main problem we are discussing here seems to be with systems. I don't know how many library catalogs can accept formats other than MARC21. Even in the open-source Koha catalog, at the time of set up you have to choose your format (Unimarc, Marc21, or something else) and stay with it. Of course, you can convert anything to the format of your choice using XML and XSL transformations, and I believe the Koha community is trying to include this in a future version. I worked with one catalog that could accept both, but it converted the records into a local format that were neither Marc21 nor Unimarc, so the original formats were lost and exporting was a disaster. If I am not mistaken, Drupal can import various formats, including MARC (in its XML version, of course). See http://drupal.org/project/marc, http://drupal.org/project/feeds_oai_pmh. (As a shameless self-advertisement, my next podcast--part 2 of open archives--will discuss some of these matters)
MARC as it is now is used for one thing only: to transfer library records among libraries. But yes, Bernhard is absolutely right: MARC-as-it-is-now will continue for a long time since with proprietary systems since it costs to upgrade. This is similar to the fact that not everyone uses the latest version of Windows because it costs money, while Linux is a different matter. (I wonder how many are still using Windows 95 or earlier versions?) As open source catalogs gain more influence, this may change however.
Concerning the 34x fields, it seems as if nobody wants to display them and I agree. Their only use would be for some kind of limits to searches, but it always seemed to me that the current records could probably have supplied essentially the same functions.