Reading through Offchain Sealing Market Design Sketch https://www.notion.so/pl-strflt/Off-chain-sealing-market-design-sketch-e145653ead9d4f07a013eac3fca3f415 there is no mention of a related idea that broadens business case for CP and addresses retrieval costs. So I’m writing it up here.
AFAIK the expensive parts of sealing and unsealing are exactly the same algorithm and require the exact same hardware configurations. CPs that are good sealers are also good unsealers.
The immediate implication is that SPs could outsource unsealing to CPs in the same way sealing is proposed, which is nice.
But here’s where it gets interesting: the SP can be removed from the unsealing picture entirely! Unsealing only requires a sealed replica and public inputs available from the chain. A retrieval client could ask for a sealed copy from their SP, get it and then send this sealed copy to a CP. The SP’s retrieval burden just went down from an hour long expensive process to a seconds long data transfer which means the clients’ likelihood of extracting the data from the busy, optimized-for-adding-sectors SP just went up. At the same time the CP’s business case just doubled. Because of specialization it’s possible the client’s total storage + retrieval costs are lower.
An interesting direction for applications built on filecoin is the use of sealed data directly in applications. Sealed data can always be decoded, it is just expensive to do so. So once applications get the sealed data from an SP they have a strong availability guarantee, only a cost/latency barrier to reading data. If applications can be structured such that latency needs are not high AND there exists an easily available CP market for doing unsealing then a cache hit into sealed data can be automatically resolved by sending it over to a CP who has a business relationship with the application.
Let’s assume the CP unseal and SP store and retrieve sealed copy functions become reliable. This is not impossible to bootstrap: we’re already on this path for SPs. If there are real efforts to making an off-chain sealing market then CP unseal capabilities would be build up as a side effect of build up in CP sealing capabilities. In this case imagine there is an application with lots of storage costs, strong availability requirements and low latency requirements. It seems like filecoin might actually be a competitive platform for building this application.