Discussion needed: Signature Creation Date in IDB Header is 3 bytes, not 4

[alexmanzer]

Referring to:

vdstools/src/commonMain/kotlin/de/tsenger/vdstools/idb/IdbHeader.kt

Line 80 in 1b449ab

val signatureCreationDate: ByteArray = rawData.readByteArray(4)

Summary
The Signature Creation Date field in the IDB Header is 3 bytes, as explicitly defined in Table 2. While Section 2.1 defines general Date encoding as 4 bytes (including a date mask), the Header specification intentionally omits the mask byte.

Rationale

ICAO Spec:
https://www.icao.int/sites/default/files/TRIP/Publications/ICAO-TR-ICAO-Datastructure-for-Barcode.pdf

1. Table 2 explicitly specifies 3 bytes

Table 2 (page 10) is the authoritative definition for the IDB Header structure:

Start	Length	Content
0x08	3	Signature Creation Date
Total header length: 11 bytes

2. The date mask is unnecessary for Signature Creation Date

Section 2.1 defines the date mask to handle partially unknown dates (e.g., a date of birth where the month is unknown). The mask indicates which parts are known vs unknown.

However, for Signature Creation Date:
• The signature creation date is always fully known – you cannot sign something on an unknown date
• A fully known date has mask 0x00 (Section 2.1: "A fully known date SHALL have a date mask of 0x00")
• Including a byte that is always 0x00 wastes space in a size-constrained barcode

3. Header is optimized for minimal size

The IDB format is designed for barcodes where every byte matters. Omitting a constant 0x00 mask byte is a sensible optimization for a field that can never have unknown components.

4. Section 3.2.4 reference to 2.1 describes the encoding method, not length

Section 3.2.4 says "Refer to section 2.1 for the encoding of this field" – this refers to how the date value is calculated (MMDDYYYY → integer → 3 bytes), not that the mask byte must be included.

Conclusion
The Header uses a compact 3-byte date representation (value only, no mask) because the Signature Creation Date is inherently fully known.

Unfortunately, in the wilderness there are both IDBs that also take 4 bytes for Signature Creation Date (Iceland), but also countries that do it right and take 3 bytes (Cameroon).
So here the ICAO Special was too indistinctible.

[tsenger]

Hi Alex, thanks for pointing that out.

You absolutely have a point. I also stumbled across the discrepancy between the length in Table 2 and the reference to Section 2.1, where the 4 bytes for the masked date are described. In a newer draft 1.20 from August 2024, there is a correction regarding the signature date:
“Corrected error in signature creation date in section 3.1.4. Should be 4 bytes including the date mask instead of 3.”

Figure 2 has also been adjusted accordingly and now specifies 4 bytes for the signature date.

But I agree with you that it actually makes no sense to use a masked signature date, as it is clearly defined. Perhaps this point could be raised again with the ICAO before a new version is published.
I will ask my colleagues to look into it.