feat: providers result limits for JSON and NDJSON#149
Draft
lidel wants to merge 2 commits into
Draft
Conversation
JSON delegated routing responses under-delivered: boxo's DefaultRecordsLimit of 20 acted as a pre-filter, and cacheFallbackIter silently dropped records without addresses. Production at delegated-ipfs.dev returned ~4 providers on JSON vs ~17 on NDJSON for the same CID. - raise the JSON cap to 100 and add a streaming cap of 1000, matching HTTP Routing v1 section 4.1.5; expose both as `--records-limit` and `--streaming-records-limit` with `SOMEGUY_*` env vars; reject negative input, 0 disables the cap - over-fetch from the underlying router by 3x (cap 3000) so that, after cacheFallbackIter drops addr-less records, the surfaced count is close to the caller's limit; cap the final stream via a new limitedIter wrapper
Member
Author
|
Triage note:
|
operators can now set --records-limit / --streaming-records-limit to any positive value; previously the multiplied request to the underlying router was clamped at 3000, capping the surviving count below high user limits. passing 0 still disables both the cap and the over-fetch. - drop cacheFallbackOverfetchMax; overfetchLimit returns limit * multiplier for positive limits - rewrite over-fetch comments: the limit is an early-termination hint, not a DHT-walk cost amplifier (per review) - slim CHANGELOG: move "callers get close to --records-limit" to Fixed, drop the over-fetch mechanism from Changed
Member
Author
|
@guillaumemichel applied feedback, but also did some thinking if doing this only in someguy is the right place (overfetch feels wasteful). Alternative approach in: |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Someguy supports JSON (buffered) and NDJSON (streaming) response types.
Problem
JSON delegated routing responses under-delivered: iiuc boxo's
DefaultRecordsLimitof 20 acted as a pre-filter, and cacheFallbackIter silently dropped records without addresses.Anecdotally (cc @aschmahmann), production at
delegated-ipfs.devreturned ~4 providers on JSON (when opened in browser) vs ~17 on NDJSON for the same CID (/ipns/ipfs.tech).As a side problem, NDJSON stream had no cap, which is a bad default.
Fix (This PR)
--records-limitand--streaming-records-limitwithSOMEGUY_*env vars; reject negative input, 0 disables the capAlternatives
Note
#149 and #150 fix the same bug two ways. Merge only one.
cachedRouterover-fetches 3x to offset records dropped for missing addresses, then caps the result. Self-contained, but 3x is a guess and the fix stays in someguy.filter-addrs/filter-protocolsunder-delivery for all consumers.Feedback welcome, but my gut feeling is the #150 is preferable: it fixes the root cause in the shared layer. Needs ipfs/boxo#1157 first.