Dickimaw Books Blog

The Problem of Name Clashes and Lost Resources for Digital Historians (RISC OS ARMTeX) 🔗

When I first saw armtex pop up on the CTAN announcements, it piqued my interest and triggered a moment of nostalgia because the first TeX distribution that I installed, back in 1992, was called ARMTeX. However, it turned out that the ArmTeX in the CTAN announcement (a system for writing Armenian with TeX and LaTeX, dating back to 1997) is completely unrelated to ARMTeX (a TeX distribution for the Acorn ARM powered computers).

There are a growing number of digital historians who are interested in documenting old computing systems from the twentieth century, but much of the information has been lost. The Internet Archive’s Wayback Machine goes back to around 1995. For reference, the World Wide Web was created in 1989, Usenet in 1980, and FTP in 1971.

Many of the operating systems developed last century had limited resources and there were restrictions on the maximum length of filenames and what characters were permitted. With short names and search engines in their infancy, there are likely many name clashes from that era, especially where the context was restricted to geographical area or operating system. This makes it quite hard for historians to search for an old application. In fact, searching for armtex yields some completely unrelated businesses.

The ARMTeX distribution that I installed on my Acorn Archimedes in 1992 (which had LaTeX 2.09) came with a DVI viewer called !DVIview, and here we have a similar problem because searching for dviview shows another application with the same name.

ARM is a RISC (Reduced Instruction Set Computer) architecture that was designed by Acorn Computers. The ARM acronym originally stood for Acorn RISC Machine and then Advanced RISC Machine. ARM split off from Acorn and is now ubiquitous in mobile devices. ARM is now commonly stylised in lowercase but here I’m using ARMTeX in order to distinguish it from the other ArmTeX. The Acorn Archimedes family of computers used Acorn’s ARM processors and ran RISC OS which replaced the earlier Arthur operating system.

Searching for RISC OS can produce results for the unrelated MIPS RISC/os UNIX operating system and the RISC-V architecture. With so many RISC-based architectures and operating systems, it’s not surprising that searching for RISC OS can end up with many false leads.

I used ARMTeX to write my PhD thesis, articles, slides for presentations, and early drafts of my novels. I started my PhD in 1991, but it was likely not until 1993 that I installed ARMTeX. (I know I used it for my first conference paper, published in August 1993.) I had first tried some word processors but I’d found them frustrating, particularly for the pages of mathematical workings that I needed to include in my thesis.

In addition to the usual advantages of LaTeX that are often cited (such as the typesetting quality and automation of cross-references and table of contents), there were two particular advantages over the word processors I had tried that are no longer as relevant to me today, given the advances in hardware. The first was the ability to split the document into separate chapters that could be included or skipped when working on a particular chapter, allowing a faster build speed. (I know this is a feature that is still used today, and I will still have chapters in separate files for very large or shared documents, but processors are significantly faster these days.)

The second advantage may seem a bit strange. Back then on RISC OS (and certain other operating systems) there was no “undelete” file function (no waste-basket, recycle bin or whatever) nor did I have an “are you sure you want to delete” confirmation prompt. (I’ve only just found out that apparently this was a configurable setting that I hadn’t enabled. 🤦) I learnt to backup the hard way. It’s a bit of a heart-stopping moment when you realise you’ve just accidentally deleted the directory containing your PhD thesis. It wasn’t even the first time something like that had happened, as a draft of one of my novels (an early version of The Fourth Protectorate) had gone the same way. In both cases, the file content was retrieved by dumping the entire free space of the hard disc, filtering out all control codes and sifting through the remaining plain text manually. This was feasible for a 32Mb drive, but for a modern 1Tb or 2Tb drive a recovery tool would be needed instead. The incident with the novel was when I was still trying a word processor and all the formatting was lost when the control codes were stripped. However, the thesis was in LaTeX so all the document source code was recovered (but not the images). It’s not an advantage that I tout these days. I now use version control (and the plain text document source lends itself well to this).

This first installation of ARMTeX had LaTeX 2.09, which meant there were no cls class files. Instead of the newer LaTeX2e \documentclass and \usepackage commands, it had the old \documentstyle command with just sty style files. The main style file was listed in the mandatory argument and the supplementary style files were listed in the optional argument.

(On the flip side, although discovering old pages is an advantage for archivists, it can be problematic if you’re searching for information on how to use the latest version of something. If you come across a LaTeX tutorial that has \documentstyle in the examples, then it’s very out-of-date.)

There were many competing operating systems in the twentieth century, each with their own idiosyncratic filing system. DOS used a backslash \ as the directory divider, and filenames were case-insensitive (but usually displayed in all-caps) and restricted to a maximum of eight characters before a dot and three after. (For example, an executable file might be called LATEX.EXE.) The absolute path started with the drive name followed by a colon. (For example, C:\MYFILES\THESIS.TEX may be a plain TeX or LaTeX file.)

The Unix-like systems have a forward slash / as the directory divider and file permissions determine whether or not the file can be run. (For example, a file called latex with the executable bit set is an application.) The filenames are case-sensitive and the absolute path starts with a forward slash. (For example, /home/myname/documents/thesis.tex may be a plain TeX or LaTeX file.)

On RISC OS, the directory divider is a dot and there were no file extensions. The file type was set using a numeric value or (if recognised) a textual synonym. The root directory is denoted with a dollar sign $ and an absolute path name would start with the filing system (such as ADFS) followed by two colons (::), followed by the drive and a dot (for example, 4.) and then the path starting with the root directory. (For example, ADFS::4.$.docs.thesis.tex is a file called tex which would need to have its type set to the hexadecimal value &2A8.) The maximum length of a RISC OS filename was 10 characters. (This was increased in later RISC OS releases.)

A RISC OS file is identified as an executable if it has an executable file type. For example, the Obey file type indicates a script. More complex applications are bundled in an application directory, which is identified as such by an exclamation mark at the start of the directory name. For example, the vector graphics application bundled with RISC OS is called !Draw (pronounced pling draw). If you double-click on the application directory’s icon in the filer, it will run a special !Run file inside the directory rather than opening the directory. (The directory can be opened with shift double-click instead.) The application directory also contains a file called !Sprites which provides the directory’s icon image, which distinguishes it from a normal directory. There is also a special file called !Boot inside the application directory which automatically sets up applicable environment variables.

The directory divider dot made it a bit awkward porting files over from another system. The Archimedes could read an MS-DOS formatted floppy disc but it would display the dot in file extensions as a forward slash. So the THESIS.TEX file on an MS-DOS disc would display in the RISC OS filer as THESIS/TEX. Since eight dot three filenames consist of 12 characters in total, which exceeded the (at that time) 10 character maximum, the extension part could end up truncated. For example, the MS-DOS file MYTHESIS.TEX would display as MYTHESIS/T. Either way, the extension is irrelevant to RISC OS, except when converting between filing systems.

This meant it was problematic porting applications that required file extensions. For my PhD, I implemented various algorithms in C, but I wrote, compiled and ran those applications on RISC OS. In order to deal with the problem of the dot as a directory divider, the C compiler flipped the file references round. For example, #include "mylib.h" would be interpreted as the file h.mylib, which refers to a file called mylib contained in a directory called h. There were similarly directories called c and o.

The ARMTeX distribution dealt with the problem in a different way. My LaTeX document thesis.tex was a file called tex in a directory called thesis. The ARMTeX LaTeX would then create files called dvi, log, aux, etc in that thesis directory (with the file types set appropriately). The \include commands in the tex file reference the directories that have the same parent as the main thesis directory. Each of those directories have a tex file (and will also have an aux file created in them).

Note that this was before PDFTeX and long before XeTeX and LuaTeX. The only output format was DVI. Images were included with \special. The format was the Archimedes’ native DrawFile format (created with !Draw). The !DVIview application was able to interpret these specials and render the image. The image dimensions were required in order to make sufficient space on the page for the image. For example, the following line is in my PhD thesis source code:

\drawfile{4.053885in}{2.227235in}{@.diagrams.quadsizeex}

\newcommand{\drawfile}[3]{\makebox[#1][l]{\rule[-#2]{0mm}{#2}\special{drawfile = #3}}}

ARMTeX was updated to LaTeX2e, but I was a late adopter as it was too large to install on my 32Mb hard drive. My husband bought a RiscPC in 1997 which had more space and we were able to install the new version on that, but eventually I got a Linux laptop and switched over to teTeX (and later TeX Live).

My son, Cameron, grew up with our old Acorn computers and learnt to program on his dad’s RiscPC. Cameron discovered some of my old files, including a copy of datetime.dtx and datetime.ins dated 2004, probsoln.dtx and probsoln.ins dated 2000, and csvtools.sty dated 2003 (missing its original dtx and ins files from which it was generated). There are also other personal packages that I never uploaded to CTAN, as well as the source code for various documents, including my PhD thesis and the LaTeX tutorials I taught for postgraduates and staff (from which my LaTeX series developed).

It’s unlikely that ARMTeX will ever be resurrected, but it was an interesting trip down memory lane. After reading the draft of this post, Cameron started investigating TeX on RISC OS, and he will provide a follow-up detailing his findings. (One of the things we discovered is that there is actually a copy of ARMTeX on CTAN in the obsolete/systems/acorn directory, but more on this in the next post.)