[issue/690] allow . and - in environment variable names.

Files touched:  schema/types/environment.go, schema/types/environment_test.go, spec/aci.md

Allow additional characters in the set considered valid in environment variable names.  Prompted by
appc/docker2aci#244 and rkt/rkt#3532.

schema/types/environment.go

@@ -21,7 +21,7 @@ import (
 )
 
 var (
-	envPattern = regexp.MustCompile("^[A-Za-z_][A-Za-z_0-9]*$")
+	envPattern = regexp.MustCompile("^[A-Za-z_][A-Za-z_0-9.-]*$")
 )
 
 type Environment []EnvironmentVariable

schema/types/environment_test.go

@@ -45,10 +45,10 @@ func TestEnvironmentAssertValid(t *testing.T) {
 			},
 			true,
 		},
-		// name with non [A-Za-z0-9_] should fail
+		// name with non [A-Za-z0-9_.-] should fail
 		{
 			Environment{
-				EnvironmentVariable{"VERBOSE-DEBUG", "true"},
+				EnvironmentVariable{"VERBOSE|DEBUG", "true"},
 			},
 			true,
 		},
@@ -65,6 +65,18 @@ func TestEnvironmentAssertValid(t *testing.T) {
 			},
 			false,
 		},
+		{
+			Environment{
+				EnvironmentVariable{"DEBUG.0", "true"},
+			},
+			false,
+		},
+		{
+			Environment{
+				EnvironmentVariable{"DEBUG-0", "true"},
+			},
+			false,
+		},
 	}
 	for i, test := range tests {
 		env := Environment(test.env)

spec/aci.md

@@ -234,7 +234,7 @@ JSON Schema for the Image Manifest (app image manifest, ACI manifest), conformin
         * **pre-start** - executed and must exit before the long running main **exec** binary is launched
         * **post-stop** - executed if the main **exec** process is killed. This can be used to cleanup resources in the case of clean application shutdown, but cannot be relied upon in the face of machine failure.
     * **workingDirectory** (string, optional) working directory of the launched application, relative to the application image's root (must be an absolute path, defaults to "/", ACE can override). If the directory does not exist in the application's assembled rootfs (including any dependent images and mounted volumes), the ACE must fail execution.
-    * **environment** (list of objects, optional) represents the app's environment variables (ACE can append). The listed objects must have two key-value pairs: **name** and **value**. The **name** must consist solely of letters, digits, and underscores '_' as outlined in [IEEE Std 1003.1-2001](http://pubs.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap08.html). The **value** is an arbitrary string. These values are not evaluated in any way, and no substitutions are made.
+    * **environment** (list of objects, optional) represents the app's environment variables (ACE can append). The listed objects must have two key-value pairs: **name** and **value**. The **name** must consist solely of letters, digits, and underscores '_' as outlined in [IEEE Std 1003.1-2008, 2016 Edition](http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap08.html), with practical considerations dictating that the name may also include periods '.' and hyphens '-'. The **value** is an arbitrary string. These values are not evaluated in any way, and no substitutions are made.
     * **isolators** (list of objects of type [Isolator](types.md#isolator-type), optional) list of isolation steps that SHOULD be applied to the app.
     * **mountPoints** (list of objects, optional) locations where an app is expecting external data to be mounted. The listed objects contain the following key-value pairs: the **name** indicates a label to refer to a mount point (which may be used by the executor when resolving a mount to a volume in the PodManifest), and the **path** stipulates where it is to be mounted inside the rootfs. The name is restricted to the [AC Name](types.md#ac-name-type) Type formatting. **readOnly** is a boolean indicating whether or not the mount point will be read-only (defaults to "false" if unsupplied).
     * **ports** (list of objects, optional) ports that this app will be listening on once started. This field is informational: example uses include helping users to discover the listening ports of the application, or indicating to executors ports that should be exposed on the host. This information could also optionally be used to limit the inbound connections to the container via firewall rules to only ports that are explicitly exposed. Each object can represent either a single port or a port range (contiguous set of ports).