Remove comment about self-describing multiformats

There is a description of the key that is used for signatures in
SignedObject, and its questionable whether you should want the
encryption scheme to be publicly available for EncryptedObject
This commit is contained in:
Mark Spanbroek 2021-01-07 22:54:33 +01:00 committed by markspanbroek
parent 15725b0c7b
commit 313e98a804
1 changed files with 0 additions and 3 deletions

View File

@ -32,9 +32,6 @@ Dagger. It is not meant to be a criticism of IPFS.
barter with (§3.4)
- Object pinning (§3.5.3) inevitably leads to centralized gateways to IPFS, such
as Infura and Pinata
- There are no self-describing multiformats for encryption and signing (§3.5.5),
although [multicodec](https://github.com/multiformats/multicodec/) can
probably be used here.
- IPFS uses variable size blocks instead of fixed-size chunks (§3.6), which
might make it a bit harder to add incentives and pricing
- Supporting version control directly in IPFS feels like an unnecessary