decentralized-web-node icon indicating copy to clipboard operation
decentralized-web-node copied to clipboard

Remove spec text: "... Messages MUST be committed to an IPLD DAG in a tree allocated to the DID of the owner ..."

Open thehenrytsai opened this issue 2 years ago • 3 comments

In order to enable data replication features for a Decentralized Web Node, all Messages MUST be committed to an IPLD DAG in a tree allocated to the DID of the owner after all subtrees are composed and committed. The top-level of Message objects MUST be committed as a DAG CBOR encoded object.

The above text is outdated especially the first sentence. The word "committed" is ambiguous in "... MUST be committed as a DAG CBOR", I read it as "it must be stored as DAG CBOR", but that's not cace, we only use IPLD DAG CBOR/PB for generating CIDs.

We need to update or remove this text.

thehenrytsai avatar Aug 24 '23 22:08 thehenrytsai

Thanks for pointing this out @thehenrytsai.

I think this entire Messages section needs updating, but the IPLD stuff definitely needs to be pulled and probably only referenced specifically where the CIDs are generated.

We'll try to get this updated for next call.

LiranCohen avatar Aug 24 '23 22:08 LiranCohen

What do you mean when you say we only use IPLD for generating CIDs? Are you saying we put data into IPLD format just for generating CIDs, but it doesn't matter how we actually store the data assuming we can still identify what data a CID points to?

If that's the case, it seems odd that we would go through all that work and not store the data that way. Also, if this is a correct interpretation, how do the current DWN libraries store data and make it so it can be referenced by CID?

WriterZephos avatar Aug 25 '23 17:08 WriterZephos

Discussed this at DIF:

  • Rationale for CID was for mainly enabling better content recovery over IPFS network without relying on a DWN,
  • IPLD was another benefit.

Assigning to @LiranCohen to respond in more detail.

andorsk avatar Sep 20 '23 16:09 andorsk