WG "resolution" of the versioning/forwards-compatibility issue

Folks,

At the XProc telcon today[1], we adopted most of the proposal[2] that
I made for versioning. The most substantial change that we made was to
reject the part of the proposal that allowed for unknown step types to
occur in p:when or p:try/p:group elements.

The persuasive arguments as I heard them were:

* It's not clear that the algorithm proposed would handle every
  conceivable kind of change that might occur. It's possible that some
  future change could lead to the algorithm for dealing with "invalid"
  steps to introduce unexpected or indeterminate changes to the flow
  graph. Some members of the WG fear that that may already be the
  case, the failing to do static analysis could lead to unexpected,
  "incorrect" behavior.

* What's more, it only works for pipelines where the author has
  explicitly set out to handle the versioning problem by using
  p:choose or p:try. Given that we have another mechanism, use-when,
  which can also be used to explicitly handle the versioning problem
  in a way that can't introduce any sort of indeterminacy, it seems
  unnecessary to support the more complex behavior described.

In short: an unknown step type in the XProc namespace (after use-when
processing) is a static error. Unexpected port names *are not* an error,
and are handled with the defaulting I proposed.

Other changes/decisions:

1. The version attribute is only allowed on p:library, p:declare-step,
   and p:pipeline.
2. Pipelines are no longer allowed to import declarations for step types
   in the XProc namespace
3. The use-when attribute is just an ordinary attribute with no special
   semantics when it occurs as the descendant of p:inline

The editor has been directed to produce a new draft which reflects
these decisions. If accepted by the WG at the next meeting, it will
become the official position of the WG and will be published as a new
*LAST CALL* working draft :-(. We'll go through the shortest possible
LC and CR periods allowed. Readers are discouraged, but cannot be
forbidden, from commenting on decisions not explicitly reopened by
this LC draft.

                                        Be seeing you,
                                          norm

[1] http://www.w3.org/XML/XProc/2009/10/22-minutes
[2] http://lists.w3.org/Archives/Public/public-xml-processing-model-comments/2009Oct/0074.html

-- 
Norman Walsh <ndw@nwalsh.com> | Mankind are always happy for having
http://nwalsh.com/            | been happy; so that if you make them
                              | happy now, you make them happy twenty
                              | years hence by the memory of
                              | it.--Sydney Smith