Avalanche interoperability of course exists between subnets, but what about on- and off-ramps to the Avalanche ecosystem? Axelar CEO Sergey Gorbunov joined other interoperability experts for a panel on interoperability at the Avalanche Summit in Barcelona, Tuesday.

Also on the panel were Billy Rennekamp of Interchain Foundation (Cosmos), Bryan Pellegrino of Layerzero and Michael Kaplan of Avalanche.

Beyond Avalanche interoperability internal to its ecosystem, the panel addressed broader needs and challenges in interoperability across Web3, including:

• The multichain future.
• Why interoperability is needed.
• Security challenges in interoperability, and where in the stack they should be addressed.
• Cross-chain topology: peer-to-peer connections versus hub-and-spoke.

The multichain future

Sergey opened the discussion with a poll of the panel: how many chains will there be? Here were the panel’s answers:

• Billy: “hundreds of thousands to millions.”
• Bryan: “thousands to tens of thousands.”
• Michael: “double digits.”

Why we need interoperability

The panel consensus converged on the need for blockchains tailored to specific use cases, and a vision for a wide diversity of those use cases. “If we’re going to see the sort of diversity of applications that need to exist, we’re going to have a need for interoperability standards for all of them,” Billy said.

More: Where will interoperability take Web3?

Cross-chain security: app layer vs. protocol

Bryan said Layerzero’s approach to security is at odds with the notion that everything has to live at the protocol layer.

“It doesn’t really make sense that there’s a world that somebody verifying on Chain B that they own a copper sword in a game on Chain A, and somebody transferring $500M, are opting into the exact same set of parameters for security,” he said. “Either somebody’s wildly overpaying or somebody’s not getting nearly enough security. For us, it’s fundamental to move all that up the stack.”

In response, Sergey raised the point that moving security decisions to the application layer raises the likelihood that developers will make mistakes in security. “Developers are not used to thinking securitywise. They want to be given https as a protocol and just run with it, with as few parameters as possible,” he said.

Bryan characterized Layerzero’s approach to application-layer security parameterization as “can vs. must.” For many categories, such as finance, one-size-fits-all may be the best approach, and many application developers will go with default settings.

Interop topology: p2p vs. hub-and-spoke

At the blockchain infrastructure level, the panel debated tradeoffs in cross-chain infrastructure, between peer-to-peer networks, where many hops are required, and hub-and-spoke networks, where honeypots may emerge.

“If you look at even how the internet shaped up. It’s not all peer-to-peer connectivity,” Sergey said. “We have peer-to-peer connections, but they’re delivered across sometimes peer-to-peer channels, sometimes hubs. … I guess the big difference that I like to think about when comparing internet topology with blockchain topology is that the more hops you take on the internet, you’re still fine, in some sense: internet only provides liveness guarantees. Deliver a packet, yes or no? All the security comes in protocols on top of it. Whereas in the blockchain space whenever we’re talking about multiple hops and if everything is completely peer to peer, then unless the network can make hundreds of thousands of connections on its own, it’s going to have to go through some intermediary connector anyways. You’re going to be assuming the security assumptions of that.”

Billy described one of the trade-offs of p2p as a loss of fungibility: “Assets that go through these different routes, they are dependent on that route,” he said. “They’re nonfungible with assets that came through other routes, because they’re dependent on being able to come back through that original route.”

More: Read about Axelar’s tech stack.

Interop security: insurance funds

As a final topic, the panel discussed insurance funds as a way to address the growing pile of losses to interoperability attacks.

“I personally would rather focus my energy on preventing that from happening rather than mitigating the damage if they do occur,” Michael said.

“You’re reducing the space of the tail [with insurance], but the tail is still completely catastrophic if you haven’t addressed it,” Bryan said.

The full video is available for now as part of the subnet stage livestream at avalanchesummit.com; the interoperability panel begins at 2h 44m.