On the 49th anniversary of documenting TCP/IP

By on 5 Dec 2022

Category: Tech matters

Tags: ,

Blog home

Remember, remember, the 5th of November: Gunpowder, treason and plot!

English Folk Verse (c.1870).

The IETF met in London for its 115th meeting and doors opened to the hackathon and pre-event semi-formal meetings on 5 November 2022. A traditionally interesting date in Britain (that’s now shed its unfortunate anti-establishment qualities) ‘Bonfire night’ is simply a time to celebrate all that’s the best of British autumn — bonfires, fireworks and food. However, in times past it was more overtly “… gunpowder, treason and plot”.

Oddly enough, the IETF meeting was held within two streets of the ‘Cato Street conspiracy’ off Edgware Road, perhaps the most incompetent revolutionary movement in the UK’s history after the Gunpowder Plot.  

Something else that was revolutionary (but not incompetent!) also celebrates a birthday of sorts on 5 November. As noted in this recent tweet from Jeff Tantsura:

Vint Cerf and Bob Kahn’s seminal paper on IEEE transactions on communications was received on 5 November 1973 and published in May 1974. It was the public release of Transmission Control Protocol (TCP) without any explicit mention of the Internet Protocol (IP) to which it is now bound (for a good reason; they hadn’t invented it yet!).

The work had been done during the months preceding, so the actual ‘birthday’ of TCP is perhaps less clear, but 5 November is minted into this paper.

The paper includes many statements of intent that go to the core motivations for a transmission layer model, the establishment of a reliable byte stream preserving sequence between two communicating entities (here called HOSTS), and pays tribute to the work of others in this space (notably Louis Pouzin whose seminal work on packet networks lies at the ground floor of most of the subsequent history).

The paper explicitly refers to CYCLADES, the network that Pouzin and others worked on to great effect, and ARPANET, which in these times was still running Network Control Protocol (NCP). The cutover to TCP/IP didn’t happen until 1983. This speaks to how much work lay ahead because ‘TCP’ of this time wasn’t adequate to the foreseen needs of a world-scale network. It was TCP, but not entirely as it’s currently known.

The paper discusses concepts such as the packet ‘header’ and the need for ‘ports’ to identify sub-contexts but also explicitly promotes fixed-format address interpretation in the header to direct routing. Remember, this is from a time before routing algorithms like Border Gateway Protocol (BGP), and so doesn’t draw on a network that dynamically informs itself of how to find things. It instead has a proscriptive structure in addressing to denote intermediate paths and elements of the path. These are ideas that made sense at the time but did not carry forward, in part because TCP without an IP layer is incomplete. When an IP layer emerged, considerations of addressing and routing moved from TCP.

In effect, this is still TCP as a work in progress. It’s not the TCP known today, but the fundamentals of its role in the delivery of bytes in packets, sequencing, and ultimately the state sequences that drive a streaming service are there.

As ARPANET expanded into a service role, the beloved applications from that time (such as FTP, TELNET, and SMTP) were ported from NCP to TCP which provided for a move to IP as the fundamental ‘substrate’ of the network. In that sense, the ‘real’ birthday of note here is 1 January 1983.

But there we were, at the London IETF meeting, on 5 November 2022. The gunpowder plot isn’t one of the most glorious memories of British history, but this other event on the 5th of November is well worth celebrating.

Rate this article

The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.

Leave a Reply

Your email address will not be published. Required fields are marked *