Internet standards are becoming more complex

As the Internet develops, it's natural to ask whether it's getting easier or harder to develop new standards, and whether the complexity of those standards is changing over time. In the following, we show that Internet standards documents have become more densely interconnected over the years, and are taking longer to publish. The IETF is becoming slower over time, because new standards are harder to develop when they must account for, and retain backwards compatibility with, increasing amounts of prior work.

The figures below show changes in the time taken from the submission of the first draft of a document through to its publication as an RFC (first plot; left), and the number of intermediate drafts each document goes through prior to publication (second plot; right). The two figures show a clear trend: documents are taking longer to make their way through the standardisation and publication process. The average time to publication was 469 days (~15 months) in 2001, rising to 1170 days (~38 months) in 2020. The number of revisions each document went though more than doubled in the same time period. The time to publish and the number of drafts prior to publication are strongly correlated, suggesting that the time is spent making changes to the document. This may go some way towards explaining the decline in output of the IETF: each document is taking longer to produce, with more revisions before publication.

This raises the question of what is causing the slowdown in RFC publication? Why is the IETF getting slower at publishing documents over time?

One possibility is that the standards documents are getting longer over time. If the length of documents is increasing, perhaps because more recent standards are trying to tackle more ambitious problems than earlier documents, then one might expect them to take longer to prepare. However, as the figure below shows, the length of documents has not changed significantly over the years.

So what has changed in the documents submitted to IETF? Two closely related factors seems to illustrate the change most clearly. The first, shown in the plot on the left below, is that there is a significant increase in the number of citations and cross references between documents: work submitted to the IETF recently is much more likely to refer to prior work. Similarly, as shown in the figure on the right below, the proportion of documents published each year that update (i.e., extend or augment) or obsolete (i.e., replace) one or more previously published documents has increased over time. In 2020, more then 30% of documents published updated or made obsolete a previously published document in the RFC series.

The next figure further confirms the growing complexity, showing how the use of keywords that indicate the normative requirements an RFC imposes on implementations has evolved over time. An early IETF document, RFC 2119, defines specific meanings for the keywords to indicate requirements on conforming implementations of the standard. The plot shows how the number of such requirements per page of each document has increased over time.

The trend is clear: as the Internet evolves, as the technical standards become more mature, and as the number of published standards increases, the time taken to develop new standards also increases. Applications and protocol standards must be maintained, and new standards need to interoperate, and build on, prior work. Dealing with this legacy, extending the Internet without breaking backwards compatibility, takes an increasing amount of time.

To learn more about this topic, please see our paper Characterising the IETF Through the Lens of RFC Deployment, presented at the ACM Internet Measurement Conference on 2nd November 2021.