Thursday, November 16, 2017

Autumn 2017 report to the Computer Conservation Society

This was presented on November 16 to the Computer Conservation Society by Doron Swade.

Work has continued steadily going through Babbage’s manuscript Sketchbooks (his working notebooks) and building up the cross reference database. As reported previously the volume of material is substantial and its organisation is not conducive to systematic study: it not indexed or themed by topic, nor is it rigorously chronological. Time and time again Babbage revisits the same or related topics over a period of decades, and these entries, many of which are cryptic, are dispersed through the twenty-five volumes comprising some 7-8,000 manuscript sheets. The cross-referenced database is being developed by Tim Robinson as a key research tool to manage this distributed content. 

The focus for the last six months has been on the volumes most relevant to the Analytical Engine design started in 1834. Happily this main sequence of books is, for the most part, in chronological order. We have just completed Volume 5 which covers 1841–1844, and includes the most intense period of work by Babbage on Plan 28, the most advanced design for the Analytical Engine.

There are notable gaps corresponding to periods when Babbage was travelling. There is a separate "Travelling Scribbling Book", which has not yet been studied in detail, and we know of a significant body of manuscript material in the Buxton archive at Oxford which was written while Babbage was travelling in Europe in the early 1840s. The intention is to fill these gaps in due course. 

While focussing on building the database, we have resisted the temptation analyse the material as we go as such analysis will be substantially assisted once the full corpus is in searchable form. Having said which, there are already new findings. A landmark drawing is Plan 25, dated 5 August 1840, which depicts the "Great Engine". This Plan provides the most complete system overview of the state of play at that time and is the most well-known of Babbage’s technical drawings. The version of the machine in Plan 25 is that Babbage presented in Turin in 1840 and on which Menabrea based his published description, which in turn prompted Ada Lovelace's famous Notes. The new finding is that there is no evidence of "user level" conditional control in this defining version of the design, this despite provision for sophisticated multi-way conditional branching at the "microcode" level. Given that the 1840 version was incapable of user-programmable conditional branching, the machine described in Plan 25 should be regarded technically as a calculator rather than a computer in the modern sense.

The first evidence of user-programmable conditional branching appears in an entry dated 19 March 1842 in what Babbage called the "minor operations" – ascertaining if a variable is zero, and ascertaining if a variable is + or -. This was when Babbage was well into the design of the "Small Analytical Engine" – a reduced version described in Plans 26 and 27. Work on Plan 28 (yet another complete reset) started two months later, but these “minor operations”, essential to making a "universal" machine, were carried over unchanged. It is yet to be established what suddenly triggered the addition of these operations. Babbage himself gives no hint of the reason in the Sketchbooks studied so far. 

Some 2,100 pages have so far been cross referenced, indexed, and transcribed into the database. Though this represents only about a quarter of the total manuscript material, we are much further than a quarter of the way through the process: most of the remaining volumes are more fragmentary, contain material not related to the Analytical Engine, or contain rough drafts later transferred to the drawings and/or the notations (textual content already captured). All the remaining volumes will be worked through though these are not regarded as a priority at this time.

Doron Swade


  1. Out of curiosity what to is used to create the "database"?

  2. Are you sure you can't share the images and crowd-source some of this work?

    1. The problem is that the s ience museum won't release scans at sufficiently high resolution to be useful. The Plan 28 group have high res scans but are contractually prohibited from sharing them. It's a VERY annoying situation because there would be a vast army of volunteers willing to pitch in if only we were given the chance.

    2. The Science Museum has put up zoomable scans here:

  3. Is the database publicly available?

  4. Plan 28 does not have copyright on the images and cannot distribute them. Copyright is held by the Science Museum and they are available to view on the Science Museum's web site.

  5. The database is not currently available but will be.

  6. How can we donate to this effort? Would love to be able to help in some tangible way.

  7. Despite having no conditional (or even unconditional) branch instruction, I have successfully demonstrated that such a machine can still be Turing complete if it's instruction cards can be looped in the manner of a Jacquard loom. I can explain how if anyone has doubts on this matter.