Added a required input and output format for workflow engines.

CHANGELOG.md

@@ -20,6 +20,8 @@ Keep the changelog pleasant to read in the text editor:
 
 version 2.0.0
 ---------------------------
++ Added a required input and output format for workflow engines.
+  [PR 357](https://github.com/openwdl/wdl/pull/357)
 
 + The input specification has been clarified.
   [PR 314](https://github.com/openwdl/wdl/pull/314) by @geoffjentry.

versions/development/SPEC.md

@@ -2896,9 +2896,15 @@ Note that because some call inputs are left unsatisfied, this workflow could not
 
 Once workflow inputs are computed (see previous section), the value for each of the fully-qualified names needs to be specified per invocation of the workflow. The format of workflow inputs is implementation specific.
 
-### Cromwell-style Inputs
+### Common WDL JSON Input Format
 
-The "Cromwell-style" input format is widely supported by WDL implementations and recommended for portability purposes. In the Cromwell-style format, workflow inputs are specified as key/value pairs in JSON or YAML. The mapping to WDL values is codified in the [serialization of task inputs](#serialization-of-task-inputs) section.
+The Common WDL JSON input format is an input format that is required to be 
+accepted by any workflow engine. It is the input format that is used for 
+performing language compliance tests on the engine.
+
+In this common format, workflow inputs are 
+specified as key/value pairs in JSON. The mapping to WDL values is codified in 
+the [serialization of task inputs](#serialization-of-task-inputs) section.
 
 In JSON, the inputs to the workflow in the previous section might be:
 
@@ -2917,7 +2923,89 @@ In JSON, the inputs to the workflow in the previous section might be:
 }
 ```
 
-It's important to note that the type in JSON must be coercible to the WDL type.  For example `wf.int_val` expects an integer, but if we specified it in JSON as `"wf.int_val": "three"`, this coercion from string to integer is not valid and would result in a coercion error.  See the section on [Type Coercion](#type-coercion) for more details.
+It's important to note that the type in JSON must be coercible to the WDL type. 
+For example `wf.int_val` expects an integer, but if we specified it in JSON as 
+`"wf.int_val": "three"`, this coercion from string to integer is not valid and 
+would result in a coercion error.  See the section on 
+[Type Coercion](#type-coercion) for more details.
+
+### Other input formats
+
+Next to the Common WDL JSON input format, engines can support any other format.
+The only requirement is that they provide a tool or documentation to convert 
+these custom formats into Common WDL JSON input format, so users can share their
+workflow inputs across the WDL ecosystem.
+
+# Computing outputs
+
+## Computing workflow outputs
+
+Every engine should have an option to provide the outputs. Either as a file,
+an API, or as stdout. The output format should be composed according to the 
+following rules:
+
+* Workflow outputs are defined in the output section of the workflow. 
+* A workflow without an output section has no outputs.
+* An engine may optionally provide outputs not listed in the output section
+  (i.e. task and subworkflow outputs) for debugging purposes, if explicitly
+  requested by the user.
+
+The above rules only apply to output formats delivered by the engine. Listing
+subworkflow and task outputs in logs is not covered by these rules.
+
+## Output formats
+
+### Common WDL JSON Output Format
+
+Every workflow engine should provide an option to provide the Common WDL JSON
+output format. This output will be used for performing language compliance tests
+on the engine.
+
+In Common WDL JSON output format the outputs are specified as key/value pairs in
+JSON. Only outputs from the workflow are provided. Sub-workflow and task outputs
+are not provided.
+
+The output value will be fully qualified. Check the [chapter on fully qualified 
+names](#fully-qualified-names--namespaced-identifiers) for more details.
+
+WDL values will be coerced to their corresponding JSON types as described in 
+the chapter on [type coercion](#type-coercion). 
+
+Example for the following workflow:
+
+
+```WDL
+workflow example {
+    ...
+
+    output {
+        String foo = cafetaria.inn
+        File analysis_results = analysis.results
+        Int read_count = readcounter.result
+        Float kessel_run_parsecs = trip_to_space.distance
+        Boolean sample_swap_detected = array_concordance.concordant
+        Array[File] sample_variants = variant_calling.vcfs 
+        Map[String, Int] droids = escape_pod.cargo
+    }
+}
+```
+
+The Common WDL output JSON could look like this:
+```JSON
+{
+    "example.foo": "bar",
+    "example.analysis_results": "/path/to/my/analysis/results.txt",
+    "example.read_count": 50157187,
+    "example.kessel_run_parsecs": 11.98,
+    "example.sample_swap_detected": false,
+    "example.sample_variants": ["/data/patient1.vcf", "/data/patient2.vcf"],
+    "example.droids": {"C": 3, "D": 2, "P": 0, "R": 2}
+}
+```
+
+### Other output formats
+
+Any other output formats may be supported by engines.
 
 # Type Coercion