Recently, decentralized storage has become very popular, but many people are too superstitious about Arweave's claim of "one-time payment for permanent storage." Permanent sounds cool, but most data isn't worth storing for hundreds of years; that kind of high upfront cost is a nightmare for high-frequency DApps.

I recently deeply tested the Walrus testnet, and I feel its approach is much more pragmatic: it doesn't sell grand narratives but truly addresses the pain point of "decentralization + affordability + usability."

In terms of architecture, it decouples storage and consensus, leveraging the advantages of Sui, with upload speeds several orders of magnitude faster than Filecoin. Using erasure coding (Red Stuff) technology, data slices are stored in a distributed manner, and as long as some slices are present, they can be restored, unlike IPFS, which relies heavily on nodes being online; its robustness is much stronger, making it enterprise-friendly.

Compared to Celestia's Rollup DA, Walrus is more versatile, capable of storing unstructured data (NFTs, videos, frontend code, etc.), reminiscent of early S3.

The downside is that the CLI is currently too primitive, error messages are confusing, which deters non-technical users. I hope the mainnet optimizes.

Overall, lowering storage costs combined with Sui Move programmability is more imaginative than simply competing for hard drive space.

@Walrus 🦭/acc $WAL #walrus $SUI