Leave an unknown notation alone and read a receipt's TLV as the id the submit answered

src/sms-id.ts

@@ -3,18 +3,17 @@ const notations = {
 	hex: { digits: /^[0-9a-f]+$/i, prefix: '0x' },
 };
 
+const places = ['receipt', 'submitResp'] as const;
+
 /** The notation a peer writes message ids in. */
 export type SmsIdNotation = keyof typeof notations;
 
 /** The notation per place the peer writes an id. An omitted place is left as it arrived. */
-export type SmsIdFormat = {
-	receipt?: SmsIdNotation | undefined;
-	submitResp?: SmsIdNotation | undefined;
-};
+export type SmsIdFormat = Partial<Record<typeof places[number], SmsIdNotation | undefined>>;
 
 export const smsIdNotations: string[] = Object.keys(notations);
 
-export const smsIdPlaces: (keyof SmsIdFormat)[] = ['receipt', 'submitResp'];
+export const smsIdPlaces: readonly string[] = places;
 
 export function isSmsIdNotation(value: unknown): value is SmsIdNotation {
 	return typeof value === 'string' && Object.hasOwn(notations, value);
@@ -25,11 +24,10 @@ const maxIdLength = 64;
 
 /**
  * The id as a plain decimal value, so an SMSC that answers a submit in one notation and writes the
- * receipt in another still correlates. An id the notation cannot read is left as it arrived, which
- * is what leaves a `<base>-<n>` id whole for DlrMerger.
+ * receipt in another still correlates.
  */
 export function normaliseSmsId(id: string, notation: SmsIdNotation | undefined): string {
-	if (notation === undefined || id.length > maxIdLength) return id;
+	if (!isSmsIdNotation(notation) || id.length > maxIdLength) return id;
 
 	const { digits, prefix } = notations[notation];