Skip to content

IPIP-252: efficient content-routing proposal for huge IPLD trees - #252

Open
petar wants to merge 5 commits into
mainfrom
petar/cr
Open

IPIP-252: efficient content-routing proposal for huge IPLD trees#252
petar wants to merge 5 commits into
mainfrom
petar/cr

Conversation

@petar

@petar petar commented Dec 1, 2020

Copy link
Copy Markdown
Contributor

No description provided.

Comment thread CONTENT_ROUTING.md
@@ -0,0 +1,264 @@
# ![](https://img.shields.io/badge/status-wip-orange.svg?style=flat-square) Content routing

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

suggestion - start with a short recap or set of references to the status quo to orient new readers (landing on problems raises the questions "problems with what" if no background is given initially)

Comment thread CONTENT_ROUTING.md

Remark: The IPFS codebase has an interface named `ContentRouting`. However, this is an interface to low-level DHT functions rather than to a wholistic content routing system as we describe here.

For instance, when bitswap fetches the contents of a CID, first it tries to find a provider using gossip-based queries to peering bitswap nodes and, if this fails, it tries to find a provider record in the DHT. This can be viewed as using two independent content-routing systems. Let's call them DHT-based and gossip-based, respectively. The DHT-based system has high latency and high success rate. Whereas the gossip-based system has lower-latency and lower success rate. To get the best of both, the DHT-based system is used as a fallback, should the gossip-based one fail.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

the "respectively" is vice versa, right? peered bitswap nodes is gossip-based.

Comment thread CONTENT_ROUTING.md

## Lack of abstraction

"Content routing" loosely refers to the mechanism that enables some nodes to advertise the data they have, and for other nodes to discover it, by a content identifier (CID). Currently content routing is not a unified software subsystem of IPFS, instead it is a workflow (involving using the DHT and other methods) that various applications (e.g. bitswap and graphsync) in need of content routing replicate individually and vary according to their need.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggest rewording last sentence to:

Currently content routing is not a unified software subsystem of IPFS, but is instead a workflow that uses the DHT and other methods. This workflow is replicated by various applications (e.g. bitswap and graphsync) and is varied according to the application's content routing needs.

@petar
petar requested a review from willscott December 14, 2020 18:31
@petar petar changed the title first draft of new content-routing spec/proposal efficient content-routing proposal for huge IPLD trees Feb 5, 2021
@lidel lidel changed the title efficient content-routing proposal for huge IPLD trees IPIP: efficient content-routing proposal for huge IPLD trees Jul 4, 2022
@lidel lidel changed the title IPIP: efficient content-routing proposal for huge IPLD trees IPIP-252: efficient content-routing proposal for huge IPLD trees Oct 26, 2022
Faramarz336

This comment was marked as spam.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: Needs Triage

Development

Successfully merging this pull request may close these issues.

4 participants