CCF EoT Analysis 2024

The End of Term Web Archive 2024 crawl is being mostly conducted by the Internet Archive using the Heritrix crawler.

This repo contains tools that ingest Heritrix logs, aggregates them at a web host level, and joins them with Common Crawl's host-level database ranking information and crawl counts.

The resulting dataset can be used to answer questions like:

• Is the crawl mostly focused on the intended US federal .gov and .mil domains?
• Are there important .gov and .mil websites that appear under-crawled?
• Are there important .gov and .mil websites that are blocking the crawler?

Makefile

Please see the Makefile for an idea of how to run things. It both builds the dataset and shows example SQL queries against it.

Dataset

If you're involved with the EoT 2024 collaboration, you can ask for permission to see or download the log dataset on Hugging Face 🤗. Please contact Greg directly on Slack or email, because we're getting a lot of random requests.

You can also Greg to have the dataset emailed, it's just 12 megabytes

The dataset is also available as a csv imported into a Google Sheet -- a link for this is posted on Slack.

Schema -- host-level db

• surt_host_name, e.g. 'com.commoncrawl'

• url_host_name, e.g. 'commoncrawl.org'

• url_host_tld, e.g. 'org'

• url_host_registered_domain, e.g. 'commoncrawl.org' for host name index.commoncrawl.org

• fed_csv_registered_domain, 0, or 1 for gov websites if in the federal .gov csv list

• hcrank -- harmonic centrality rank from the web graph

• prank -- page rank from the web graph

• ccf_captures -- the count of webpage captures for this host in CC-MAIN-2024-33 (just 200s)

• captures -- the count of webpage captures for this host in EoT 2024 (all codes, not just 200)

• robots_duration -- milliseconds, for the most recent robots fetch (string)

• robots_code -- http code for the most recent robots fetch (string). -404 was a 200 but empty file.

HTTP response code counts (integers)

• codes_2xx # success
• codes_3xx # redir
• codes_4xx # temporary error
• codes_5xx # permanent error
• codes_0 # did not fetch for some Nutch precursor reason (dns, robots)
• codes_999 # unusual we-hate-crawler code
• codes_666 # ditto
• codes_xxx # catchall

Failed column ideas

The following column ideas came from the Nutch documentation, but turn out to never happen. We will remove them in the next iteration.

• codes_minus_4xx
• codes_other_minus
• codes_robots_disallowed
• codes_robots_fetch_failed
• codes_network_failed

TODOs

• remove zero columns

• csv equivalent to the parquet

• percentage columns, to avoid SQL arithmetic?

• mimetypes counts (grouped by text/ video/ audio/)

• domain-level rollups