When it works on the other vendor is not a bug report

When two vendors’ equipment does not interoperate, the instinct is that one of them is broken. Usually neither is. They have made different, legitimate choices where the specification left room.

Where the room is

Specifications define what must happen and leave much unstated. Timer defaults. Which optional capabilities are enabled without being asked. How to behave when a peer does something the specification calls undefined. What to do with a malformed field — drop, ignore, or attempt to interpret.

Two implementations can be fully compliant and still fail to work together, because compliance does not require identical choices.

A concrete example

OSPF requires matching MTU for an adjacency to reach full state. Vendors count MTU differently — whether the layer 2 header is included — so two devices configured to the same number may disagree about what that number means, and the adjacency stops at EXSTART. Both are behaving correctly. Both are following the standard. The adjacency does not form.

How to approach it

Stop asking which one is wrong. Ask what each is doing, from a capture rather than from documentation. The packets show what was actually sent, which is frequently not what either vendor’s documentation implies.

Then find the specific parameter they disagree about and set it explicitly on both. The fix is nearly always making an implicit value explicit at both ends.

Explicit beats default

In any multi-vendor deployment, configure explicitly the things you would leave to defaults in a single-vendor one. Timers, authentication, MTU, spanning tree mode and priority, negotiation settings. It is more configuration, and it means the behaviour is defined by your configuration rather than by two vendors’ independent decisions.

When raising a case

“It works with the other vendor” invites the answer that the other vendor is non-compliant, and you will get nowhere. “Here is a capture showing device A sending X and device B responding Y, and here is the section of the RFC I believe applies” gets a real engineer. The evidence changes the conversation entirely.