IOAM VXLAN GPE: Conversion to/from VXLAN (without GPE)

[mickeyspiegel]

In many potential VXLAN deployment scenarios, IOAM domain boundaries might not align with VTEP locations. At the IOAM encapsulating node, a packet might be received from outside of the IOAM domain, that already has a VXLAN or VXLAN-GPE encapsulation.

Given current deployment trends, it is much more likely that such a packet has a VXLAN (without GPE) encapsulation. Rather than throwing on multiple layers of VXLAN / VXLAN-GPE encasulations, it would be much more efficient for the IOAM encapsulating node to convert the existing VXLAN encapsulation to a VXLAN-GPE encapsulation.

Such an approach leads to an issue at the IOAM decapsulating node:
How can the IOAM decapsulating node determine whether to leave the VXLAN-GPE encapsulation as is, or to convert the VXLAN-GPE encapsulation back to VXLAN?

Should a bit be defined in the "reserved" part of the VXLAN header that indicates whether the original encapsulation was VXLAN or VXLAN-GPE?
This would allow the IOAM decapsulating node to determine whether it needs to convert back to a VXLAN (without GPE) encapsulation.