Out of order IDs mixed with missing IDs

[rhansen]

TAPv14 says:

Test Points may be output in any order, but any Test Point ID provided must be within the range described by the Plan.

TAPv14 also says:

If there is no number, the harness must maintain its own counter until the script supplies test numbers again.

How do these two interact? For example, how should a harness process the following document:

1..4
ok 2
ok
ok 1
ok 3

I can see several reasonable choices for the implied ID of that second test point:

• 1 because that was the lowest ID not yet affiliated with any test points when the second test point was observed
• 2 because the ID-less test point was the second test point observed
• 3 because that is one greater than the highest affiliated test point ID when the second test point was observed
• 4 because that was the only ID not affiliated with another test point at the conclusion of the TAP document

My opinion: If the N^th test point in the TAP document does not have an ID, then the harness gives it ID N even if ID N is already affiliated with another test point. Thus, the above stream would be equivalent to:

1..4
ok 2
ok 2
ok 1
ok 3

Rationale:

• It is the easiest for a harness to implement.
• A test point's ID is immediately known when the test point is observed.
• The ID is guaranteed to be within the plan's range unless there are too many test points.
• It would effectively require TAP document producers to always include the ID with their test points if the test points might be written out of order.

Thoughts?