惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Apple Machine Learning Research
Apple Machine Learning Research
Jina AI
Jina AI
博客园_首页
WordPress大学
WordPress大学
罗磊的独立博客
小众软件
小众软件
Last Week in AI
Last Week in AI
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
Hugging Face - Blog
Hugging Face - Blog
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
爱范儿
爱范儿
The Cloudflare Blog
GbyAI
GbyAI
C
Check Point Blog
腾讯CDC
MyScale Blog
MyScale Blog
有赞技术团队
有赞技术团队
博客园 - 聂微东
IT之家
IT之家
雷峰网
雷峰网
H
Help Net Security
博客园 - 叶小钗
美团技术团队
D
DataBreaches.Net

Common Crawl

Common Crawl - Blog - Introducing the AI Visibility Audit Common Crawl - Blog - Host- and Domain-Level Web Graphs March, April, and May 2026 Common Crawl - Blog - May 2026 Crawl Archive Now Available Common Crawl - Blog - April 2026 Crawl Archive Now Available in a Hugging Face Storage Bucket Common Crawl - Blog - You can now build directly on Common Crawl from the browser Common Crawl - Blog - Host- and Domain-Level Web Graphs February, March, and April 2026 Common Crawl - Blog - April 2026 Crawl Archive Now Available Common Crawl - Blog - April 2026 Common Crawl Newsletter Common Crawl - Blog - Host- and Domain-Level Web Graphs January, February, and March 2026 Common Crawl - Blog - March 2026 Crawl Archive Now Available Common Crawl - Blog - IPv6 Adoption Across the Top 100K Web Hosts Common Crawl - Blog - Web Graph Statistics Gets a Proper Upgrade Common Crawl - Blog - Measuring Web Accessibility from Crawl Archives Common Crawl - Blog - Announcing the Whirlwind Tour of Common Crawl's Datasets Using Java Common Crawl - Blog - Host- and Domain-Level Web Graphs December 2025 and January/February 2026 Common Crawl - Blog - Introducing the New Examples & Resources Browser Common Crawl - Blog - February 2026 Crawl Archive Now Available Common Crawl - Blog - AI Plumbers at FOSDEM’26 Common Crawl - Blog - CC-Citations: A Visualization of Research Papers Referencing Common Crawl Common Crawl - Blog - CommonLID: Re-evaluating State-of-the-Art Language Identification Performance on Web Data Common Crawl - Blog - Host- and Domain-Level Web Graphs November/December 2025 and January 2026 Common Crawl - Blog - January 2026 Crawl Archive Now Available Common Crawl - Blog - Web Archives for Social Sciences Datathon, Bristol Common Crawl - Blog - How SEOs Are Using Common Crawl's Web Graph Data for AI Ranking Signals Common Crawl - Blog - GneissWeb Annotations Examples Common Crawl - Blog - Common Crawl at the Mozilla Festival 2025 Common Crawl - Blog - Host- and Domain-Level Web Graphs October, November, December 2025 Common Crawl - Blog - December 2025 Crawl Archive Now Available Common Crawl - Blog - A Sampling of 2025 Research Referencing Common Crawl Common Crawl - Blog - Host- and Domain-Level Web Graphs September, October, and November 2025
Common Crawl - Blog - Announcing a Change to Common Crawl...
2026-04-01 · via Common Crawl

Note

This post was published on 1 April 2026. The nibble remains a real unit. 🐇

Following an internal review of our data reporting methodology, Common Crawl is today announcing a change to the unit of measurement used when publishing dataset size figures. Effective immediately, all dataset sizes will be reported in nibbles (also written: nybbles) rather than bytes.

This decision was not taken lightly.

Background

A nibble is a well-established unit of digital information, formally defined as four bits, or one half of an octet. The nibble has a distinguished history in computing, appearing in early processor documentation, BCD encoding schemes, and hexadecimal representation contexts where its properties are particularly convenient. Despite this pedigree, the nibble has been consistently underrepresented in large-scale dataset reporting (which is an oversight we believe it is now time to correct.)

Rationale

The byte, defined as eight bits, has long dominated discourse around dataset scale. We do not dispute the byte's utility in many applied contexts. However, for the purposes of communicating the scope of Common Crawl's holdings to the research community, the nibble offers a number of meaningful advantages.

First, nibble-denominated figures provide a more granular representation of dataset scale. A figure reported in nibbles conveys the same information as one reported in bytes, while offering approximately twice the numerical precision in the sense that the numbers are approximately twice as large, and therefore easier to distinguish from one another at a glance.

Second, nibble-based reporting brings Common Crawl into closer alignment with the hexadecimal community, a constituency whose contributions to computing we have perhaps not sufficiently acknowledged in our public communications.

Third, and most significantly: the Common Crawl Foundation exists to preserve the web's content for future generations of researchers. It would be inconsistent, some might even say unconscionable, to champion the preservation of data while allowing a legitimate and historically significant unit of data measurement to quietly disappear from active use. The nibble deserves better. We intend to see that it gets it.

Our most recent crawl, previously reported at approximately 344 tebibytes, can now be accurately described as exceeding 689 tebibbles. We consider this an improvement on multiple counts.

Illustrative Comparison

The table below illustrates the effect of this change on selected historical crawl figures.

Crawl Previous Reported Size Revised Reported Size
CC-MAIN-2013-20 ~102 TiB ~204 Tib
CC-MAIN-2019-18 ~198 TiB ~396 Tib
CC-MAIN-2024-10 ~424.7 TiB ~849.4 Tib

Note

"Tib" denotes tebibbles. We acknowledge that this unit does not yet appear in ISO/IEC 80000-13 and we have submitted a request to the relevant working group.

Frequently Anticipated Questions

Will this affect the actual data?

No. The underlying corpus is unchanged. Only the reported size is affected.

Is a nibble a real unit?

Yes.

Why stop at nibbles? Why not bits?

We considered bits. The resulting figures, while impressive, were felt to risk confusion with latency measurements. Nibbles represent a pragmatic compromise between scientific rigour and legibility. A further migration to bits remains under evaluation for a future reporting cycle.

Why not octal?

Octal’s situation is understood, and we are not unsympathetic, but only luusers use octal. Octal digit boundaries fall on 3-bit groupings, which do not divide evenly into nibbles. This structural incompatibility makes octal difficult to accommodate within the current reporting framework. We encourage the octal community to seek representation through appropriate channels.

Is this commitment to nibble preservation sincere?

The Common Crawl Foundation has preserved over 300 billion web pages for the benefit of researchers worldwide. We do not undertake preservation efforts casually. The nibble is a real unit, it is in measurable decline as an active reporting convention, and we are in a position to do something about that. We have chosen to act.

Is this a joke?

We refer you to our track record of responsible stewardship of one of the web's most significant open datasets, and decline to comment further.

Conclusion

Common Crawl remains committed to transparency, open access, and the principled application of information-theoretic concepts to public data reporting. The preservation of digital heritage takes many forms. Some of them involve bytes. Going forward, more of them will involve nibbles.

Updated documentation reflecting the new unit will be published to our website in due course.