The SCORM 2004 data model: every cmi element, and what changed from 1.2

What 2004 changed

SCORM 2004 keeps the shape of SCORM 1.2 — a course finds a JavaScript object, calls functions on it, reads and writes string values — and changes almost every detail. The object has a different name. The functions have different names. The status field became two fields. Times are written in a different format. If you have read the 1.2 data model reference, the fastest way to hold both in your head is this: nothing carries over unchanged except the idea.

Editions matter. 2004 2nd Edition, 3rd Edition and 4th Edition are all "SCORM 2004" and differ in ways that bite: the suspend-data cap is smaller in 2nd Edition than in 1.2, and 4th Edition added the shared-data elements and tightened sequencing. A package declaring 2004 4th Edition to a platform that implements 3rd may import and then behave unexpectedly.

The eight API functions

Same count as 1.2, all renamed. Every argument and every return value is a string.

The 2004 error space encodes the call sequence, which makes debugging much easier than in 1.2. A failed SetValue before Initialize is 132; after Terminate it is 133. The equivalent for GetValue is 122 and 123, and for Commit, 142 and 143. If you see any of those six, the problem is when the call happened, not what it contained.

Status, score, and progress

The 2004 change people actually feel is the status split. SCORM 1.2 has one field that answers two questions badly; 2004 has two fields that answer them separately.

Objectives and interactions

Both groups are arrays, indexed from zero, addressed as cmi.objectives.0.id and so on. cmi.objectives._count and cmi.interactions._count are read-only and tell you how many exist. The index must be written in order: writing index 3 when _count is 1 is error 407.

Objectives carry their own status and score, which is how a single SCO reports on several learning objectives:

Interactions are the per-question records. The big 2004 change is that they are read/write — 1.2 made them write-only, so content could not read back what it had stored, and many platforms accepted and discarded them.

Navigation, exit, and resume

2004 lets content ask the LMS to navigate, which 1.2 could not do at all. The request is written before Terminate and acted on after it.

Whether those requests are honoured depends on the control modes in the manifest's sequencing block, which is a separate subject covered in SCORM 2004 sequencing.

Exit and resume work as in 1.2 but with an extra value and a second mechanism. cmi.exit (write-only) takes time-out, suspend, logout, normal, or an empty string; normal is new in 2004. To be resumable, an attempt must end with cmi.exit set to suspend — or, in a multi-SCO course, with adl.nav.request set to suspendAll. On the next launch, cmi.entry reads resume and cmi.suspend_data returns what was stored. Ending with an empty exit closes the attempt on many platforms and discards the suspend data, which is the most common cause of "it forgot my progress" that is not a size problem.

Error codes

All twenty-four, grouped as the specification groups them. In practice the 1xx family tells you the session is out of order and the 4xx family tells you the element or value is wrong.

Codes 403 and 405 swap meanings between 1.2 and 2004 — in 1.2, 403 is "read only" and 404 is "write only". Content ported between versions that interprets codes numerically will misreport them.

Mapping every 1.2 element to 2004

For porting content or reading a runtime log from the other version:

Porting is therefore never a search-and-replace: the status split needs a decision about what "done" and "passed" mean for your content, and the time format change breaks silently with error 406 if missed. Which version to build for a given destination is covered in SCORM 1.2 vs 2004, and when a ported package misbehaves, work through the diagnostic checklist.