Common Crawl vs The Stack v2

Operates a distributed web crawler (CCBot) that systematically traverses 3-5 billion web pages monthly, capturing raw HTML, metadata, and response headers into WARC (Web ARChive) format files stored on AWS S3. The crawl respects robots.txt directives and maintains an opt-out registry for content exclusion. Each monthly snapshot is immutable and indexed for retrieval, creating a cumulative archive of 300+ billion pages spanning 15+ years of web history.

Unique: Operates the largest open web crawl archive with 300+ billion pages spanning 15+ years, maintained as a non-profit public good with monthly refresh cycles and dual indexing (CDXJ + columnar) for both URL-based and structured queries. No commercial competitor maintains equivalent historical depth and scale.

vs alternatives: Larger, older, and more freely accessible than commercial web archives (Wayback Machine, Archive.org) with explicit support for ML training pipelines and no rate-limiting for research use.

Provides CDXJ (Capture inDeX JSON) indices that map URLs to byte offsets within WARC files, enabling direct random access to specific pages without scanning entire archives. Queries specify a URL and optional date range, returning matching captures with metadata (HTTP status, content type, timestamp). This index layer abstracts away WARC file complexity and enables efficient lookup of historical versions of individual pages.

Unique: Uses CDXJ standard (JSON-based capture index) rather than proprietary indexing, enabling interoperability with other web archive tools and allowing byte-offset-based random access to WARC files without full-file decompression. Supports both exact and wildcard URL matching.

vs alternatives: More efficient than sequential WARC scanning for URL lookups and more standardized than Wayback Machine's custom index format, enabling third-party tool integration.

Publishes infrastructure status updates, known issues, and errata for crawls through a public status page and mailing list. Issues are documented with affected crawls, impact assessment, and workarounds. Status monitoring includes S3 availability, index health, and crawl progress. Errata tracking enables users to identify and work around data quality issues in specific crawls.

Unique: Maintains public errata tracking and status monitoring for crawls, enabling users to identify and work around data quality issues. Combines status page, mailing list, and documentation for transparency.

vs alternatives: More transparent than proprietary data sources; public errata tracking enables community awareness of issues, whereas most competitors provide no visibility into data quality problems.

Operates a distributed web crawler (CCBot) that can be configured with custom crawl parameters including politeness delays, user-agent strings, robots.txt interpretation, and domain-specific crawl budgets. The crawler respects HTTP standards and robots.txt directives, with configurable behavior for handling redirects, timeouts, and errors. Crawl parameters are documented for each monthly release, enabling reproducibility and evaluation of crawl quality.

Unique: Publishes crawl parameters and methodology for each monthly release, enabling reproducibility and evaluation of crawl quality. Crawler respects HTTP standards and robots.txt, with documented politeness policies.

vs alternatives: More transparent about crawl methodology than proprietary crawlers; published parameters enable reproducibility and comparison with other crawling approaches.

Provides columnar indices (format and query syntax unspecified in documentation) that enable structured queries across archive metadata without parsing WARC files. Queries can filter by domain, content-type, HTTP status, crawl date, and other fields, returning matching page metadata and offsets. This approach trades random-access flexibility for efficient bulk filtering and aggregation across billions of pages.

Unique: Uses columnar storage (likely Parquet or similar) for metadata indices, enabling efficient filtering and aggregation across billions of pages without decompressing WARC files. Supports multi-field queries and bulk statistics generation.

vs alternatives: More efficient than CDXJ for bulk filtering and aggregation queries; enables data engineers to pre-filter before WARC parsing, reducing downstream processing costs.

Extracts hyperlink relationships from crawled pages to construct a directed web graph showing which pages link to which other pages. This graph data is provided separately from raw page content, enabling analysis of link structure, PageRank-like metrics, and domain authority without parsing HTML. The extraction process identifies both internal (same-domain) and external (cross-domain) links.

Unique: Extracts hyperlink graph from petabyte-scale web crawl, providing researchers with a snapshot of global web topology at monthly intervals. Graph data is separated from content, enabling efficient analysis without parsing HTML.

vs alternatives: Larger and more recent than academic web graph datasets (e.g., WebGraph, SNAP); freely available and updated monthly, whereas most academic graphs are static or years old.

Enables retrieval of any page version from the cumulative 300+ billion page archive spanning 2007-present, with monthly granularity. Users specify a URL and date range, and the system returns all captures of that page from matching crawls. This creates a time-series view of how individual pages evolved, including content changes, design updates, and deletion/resurrection events.

Unique: Maintains 15+ years of monthly web snapshots (300+ billion pages cumulative), enabling fine-grained temporal analysis of web content evolution. No commercial competitor offers equivalent historical depth at this scale.

vs alternatives: Larger and more comprehensive than Internet Archive's Wayback Machine for bulk historical analysis; free and designed for programmatic access rather than interactive browsing.

Exports raw web content in WARC (Web ARChive) format, a standardized container that bundles HTTP request/response pairs with metadata. Each WARC record includes the original HTTP status code, headers, response body (HTML, JSON, binary), and crawl metadata (timestamp, IP address, user-agent). WARC files are gzip-compressed and stored on S3, with indices enabling random access to specific records without decompressing entire files.

Unique: Uses WARC standard format (ISO 28500) rather than proprietary encoding, ensuring long-term preservation and interoperability with other archival tools. Stores on AWS S3 with public access, enabling direct programmatic access without intermediary APIs.

vs alternatives: More standardized and preservation-friendly than custom formats; larger and more recent than academic web corpora; free and designed for large-scale processing rather than interactive access.

Aggregates 67 TB of source code from the Software Heritage archive, filtering for permissively licensed repositories (MIT, Apache 2.0, BSD, etc.) across 600+ programming languages. Uses automated license detection and validation to ensure legal compliance for model training. Implements a rigorous deduplication pipeline at file and repository levels to eliminate redundant training data and reduce dataset bloat.

Unique: Largest open-source code dataset at 67 TB with automated opt-out governance allowing repository owners to request removal, combined with rigorous deduplication and PII removal pipeline — no other public dataset offers this scale with legal compliance and community control mechanisms

vs alternatives: Larger and more legally compliant than GitHub's CodeSearchNet (14M files) or Google's BigQuery public datasets, with explicit opt-out governance vs. implicit inclusion, and covers 600+ languages vs. Codex training data's undisclosed language distribution

Implements a community-driven opt-out system where repository owners can request removal of their code from the dataset without legal takedown notices. Maintains a registry of excluded repositories and re-applies exclusions during dataset updates. Provides transparent governance documentation and a clear submission process for removal requests, balancing open access with creator rights.

Unique: First large-scale code dataset to implement opt-out governance at dataset level rather than relying solely on license compliance, with transparent registry and community submission process — shifts power from dataset creators to code contributors

vs alternatives: More respectful of creator autonomy than GitHub Copilot's training approach (no opt-out) or academic datasets (one-time snapshot), and more scalable than individual DMCA takedowns

Automated pipeline that scans source code for personally identifiable information (email addresses, API keys, SSH keys, credit card patterns, phone numbers) and removes or redacts them before dataset release. Uses regex patterns, entropy-based detection for secrets, and heuristic rules to identify sensitive data. Operates at file level with configurable sensitivity thresholds to balance data utility against privacy risk.

Unique: Combines regex pattern matching, entropy-based secret detection, and heuristic rules in a unified pipeline with configurable sensitivity — more comprehensive than simple regex-only approaches, but trades off false positive rate against security coverage

vs alternatives: More thorough than GitHub's secret scanning (which only flags known patterns) because it includes entropy-based detection for unknown secret formats, but less accurate than specialized tools like TruffleHog due to language-agnostic approach

Indexes 67 TB of source code across 600+ programming languages with language-aware metadata (syntax, file extension, language family). Enables retrieval by language, license, repository, or code patterns. Uses Software Heritage's existing indexing infrastructure as foundation, augmented with language detection and classification. Supports both bulk download and filtered queries for specific language subsets.

Unique: Leverages Software Heritage's existing language detection and indexing infrastructure, then augments with BigCode-specific language classification and filtering — avoids reinventing language detection while providing dataset-specific query capabilities

vs alternatives: More comprehensive language coverage (600+ languages) than GitHub's Linguist (500+ languages) and more accessible than Software Heritage's raw API because it's pre-filtered for permissive licenses and deduplicated

Removes duplicate code files and repositories using content hashing (SHA-256 or similar) and fuzzy matching for near-duplicates. Operates in two stages: exact deduplication via hash matching, then fuzzy matching (e.g., Jaccard similarity or MinHash) to catch semantically identical code with minor formatting differences. Preserves one canonical copy of each unique code pattern while removing redundant training examples.

Unique: Two-stage deduplication combining exact hash matching with fuzzy similarity matching (likely MinHash or Jaccard) to catch both identical and near-identical code — more thorough than single-stage approaches but computationally expensive

vs alternatives: More aggressive deduplication than CodeSearchNet (which uses simple hash matching) because it catches near-duplicates, but less semantic than clone detection tools (which understand code structure) because it's content-based

Integrates with Software Heritage's comprehensive archive of 200+ million repositories and their full version control history. Extracts source code snapshots from Software Heritage's Git/Mercurial/SVN repositories, preserving repository metadata (commit history, author info, timestamps). Provides access to code at specific points in time, enabling historical analysis or training on code evolution patterns.

Unique: Leverages Software Heritage's universal code archive (200M+ repositories) as data source, providing access to code that would be impossible to collect via GitHub API alone — enables training on archived/deleted repositories and non-GitHub platforms (GitLab, Gitea, etc.)

vs alternatives: More comprehensive than GitHub-only datasets because it includes code from GitLab, Gitea, SourceForge, and other platforms archived by Software Heritage; more legally defensible than web scraping because it uses an established, community-maintained archive

Tracks and validates SPDX license identifiers for each repository, ensuring only permissively licensed code (MIT, Apache 2.0, BSD, etc.) is included. Maintains license metadata alongside code files, enabling downstream users to verify legal compliance. Implements license hierarchy and compatibility checking to handle dual-licensed or complex licensing scenarios.

Unique: Combines automated SPDX detection with manual review and maintains license metadata alongside code, enabling downstream users to verify compliance — more transparent than datasets that simply claim 'permissive licenses' without proof

vs alternatives: More legally rigorous than GitHub's CodeSearchNet (which doesn't validate licenses) and more transparent than Codex training data (which doesn't disclose license filtering at all)

Maintains versioned snapshots of the dataset (e.g., v2.0, v2.1) with documented changes between versions (new repositories added, deduplication improvements, PII removal updates). Provides checksums and manifests for reproducibility, enabling researchers to cite specific dataset versions and reproduce results. Tracks dataset lineage and transformation history.

Unique: Maintains semantic versioning and detailed changelogs for dataset releases, enabling researchers to cite specific versions and understand dataset evolution — more rigorous than one-off dataset releases without versioning

vs alternatives: More reproducible than academic datasets that are released once without versioning, and more transparent than commercial datasets (Codex) that don't disclose version history or changes