search

RhizomeDB / pomo-ipld

The IPLD representation of PomoDB
Creative Commons Zero v1.0 Universal
2 stars 0 forks source link

Some questions that occured to me while reading this spec #2

Open Gozala opened 10 months ago

Gozala commented 10 months ago
  1. IPLD schema says that entity id is bytes, but then int is used to in the examples. This is really confusing and I'm not sure which one out of two is correct

https://github.com/RhizomeDB/pomo-ipld/blob/65484801ef5159ae42f701895ffee25e88ebd7ef/README.md?plain=1#L51-L64

  1. Could you please provide some rational for the 128bit limit rational ? I'm under impression that perhaps use ofy u128 was intention which might explain my first question

https://github.com/RhizomeDB/pomo-ipld/blob/65484801ef5159ae42f701895ffee25e88ebd7ef/README.md?plain=1#L68

  1. Should CausedBy canonicalization be prescribed, or better yet be defined as a Map<CIDBytes, Unit> to ensure that same fact will hash the same ?

https://github.com/RhizomeDB/pomo-ipld/blob/65484801ef5159ae42f701895ffee25e88ebd7ef/README.md?plain=1#L105-L109

  1. I think something is missing in this sentence, is trivial to handle ?

the same fact entered into the store twice is trivial for operations that only depend on the graph structure of the store

  1. Perhaps it's better idea to recommend using multihashes instead ? IPFS stack through experience settled on using multihashes (stripping codec and cid tags) for block stores as that information usually does not affect most operations. In case of DB it makes little sense IMO to have even hashing algorithm retained, individual DB could probably choose whatever algorithm it cares and use untagged digests for keys. When communicating with others DB could choose to add prefixes etc...

https://github.com/RhizomeDB/pomo-ipld/blob/65484801ef5159ae42f701895ffee25e88ebd7ef/README.md?plain=1#L145-L153

Gozala commented 10 months ago

I also have been talking to @expede on discord about what were limits for causedBy and how to handle cases where you may need to have more than that number. I think spec should probably cover that

expede commented 10 months ago

Yep agreed! The spec is very WIP and Fission has it on hold. PRs welcome :)

Gozala commented 10 months ago

I'm interested in contributing, having above questions answered would enable me to make align with the vision.