At some stage Mac OS X dawned upon us. And while you could run your old TEX installations in Classic mode, it was clear that this was rather far from ideal. The Mac had gained a Unixy underbelly and thus it seemed natural to let it run a Unixy TEX as well. Luckily the Unix-wise people thought the same and TEX systems that could run in Mac OS X’s shell were available pretty soon. And using the right installers those pretty much ‘just worked’ as well.

In addition, Mac OS X’s PDF savvy display system and pdfTEX just seemed to be made for one another: Where Mac TEX developers had to spend ages writing DVI previews in Classic times, they could just use pdfTEX instead of traditional TEX now and let the OS do the hard drawing work. Brilliant.

Which suggests that people had plenty of time at their hands for creating great environments to actually use TEX in. And, sure enough, at least two such environments exist today: TEXShop and iTEXMac. Both essentially do the job (text window, syntax colouring, TEX button, BibTEX button, preview window), both could be considered posterboy open source projects (somewhat technical application, somewhat buggy, somewhat byzantine, somewhat hard to understand code aggravated by the many collaborators, somewhat lacking polish) and neither of them is really good. Personally I prefer TEXShop because there’s less clutter in it.

So while we have an excellent starting point for using TEX in OS X these days from a technical point of view, actually using TEX hasn’t improved that much. (All right it has, but only because previewers can [try to] jump back to the matching place in source code when you click there) UI-wise I keep thinking that working on projects with several files things still haven’t surpassed my old TEX shell on the Atari. And that shell being particularly powerful isn’t the reason for that. And -  another trend on Mac OS X - the stuff beneath the UI can be quite advanced and hard to understand. Hard enough to be considered unusable if you don’t want to devote a lot of time to it. In a way I find this quite frustrating because it always suggests that you can ‘in principle’ solve the problem if only you try hard enough whereas in Classic Mac OS if things didn’t work they didn’t work. No need to blame yourself, the programmers just didn’t want you to do it.

One thing that regularly drives me nuts in terms of low-level TEX on OS X is that TEX seems determined to write all its output files in the same folder that your input file lives in. I don’t appreciate that and I’d much rather have my log, and toc and possibly even PDF files stowed away in different folders than cluttering up the folder with the actually valuable source file. My study of kpathsea stuff (manual pages, argh!) suggests that this just isn’t possible. Or that my Unix-fu isn’t strong enough to bend the computer’s will. So I have given up on this, but I invite you tell me how to solve this.

I could go on for a while, writing about how TEXShop seems to be even better at stashing or leaking memory than Safari, how it degrades when run for a long time, how it manages to frequently have just the wrong ‘key’ window and thus you have to switch windows (possibly with OS X’s far-from-ideal window cycling, argh!) or even use the mouse before being able to continue.

I could also highlight other things about TEX on the Mac like BibDesk which takes the pain out of using BibTEX, like the nice services there are to magically turn a typed TEX command into a formula that will be inserted into another document as a graphic. Or I could go on raving about exciting new developments like the microtypography we see in pdftex or the path to better support for Unicode and fonts that is done by XƎTEX. TEX is very much alive and kicking on OS X but it doesn’t ‘feel’ as good as I think it should in day-to-day usage.

When getting a Mac in 1993, I obviously also wanted to use TEX on it. And while I thought the Mac was fantastic, it soon became clear that TEX and the Mac weren’t the most natural companions. While ‘lesser’ computers like the Atari or IBM compatibles could naturally run applications in text mode, the Mac’s all-graphic approach made this hard. Which in turn meant extra effort for people implementing TEX systems. Yet, there were a number of TEX implementations available for the Mac. Just that I had to decide which one I wanted.

This being in 1993 meant that I couldn’t simply invest ten minutes in downloading the different TEX implementations and trying them out but that I had to use some of my pocket money to order ten or so floppy disks to get them from my TEX user group or find people having those disks and arrange something with them. In fact my memory about the chronology of all this isn’t quite clear. But essentially there were various TEX implementations available. Direct TEX, Oz TEX and I think also CMac TEX and later on the legendary TEXtures.

DirectTEX sounded quite good to me but was quite pricey: Both because of its shareware price which required you to buy multiple licenses in one go and because it required you to have Apple’s MPW Shell which essentially provided a command line ‘development’ environment on the Mac in which the various tools of a TEX installation could run. It was hideously expensive, though, either by itself or by requiring you to be a member of Apple’s pricey developer programme to buy it. So it was a bit out of range.

OzTEX was quite a bit more Mac like in that it unified running TEX and displaying the previews into a single application (with a separate OzMF existing in case you needed to create new font bitmaps). Just by the look of its cute EX lion icon ( which I use to this day for my modified EX keyboard layout) I liked it right away. The fact that DANTE bought a license for all its members also made using OzTEX very attractive. The only problem with it remaining that I never really liked its way of working for some reason.

I think I tried CMacTEX once but it never much resonated with me and I was drooling when I first saw TEXtures - an application that was so cool and pricey that it was beyond belief: The pulled the trick of continuously running TEX while you typed and displaying TEX’s output ‘almost live’ in a second window. All that back in the mid 1990s.

While today you can just blindly process pretty much any TEX document in a few seconds, back then seeing TEXtures was amazing. They must have optimised things rather well for the results on the machines back then to be acceptable. They also created outline PostScript versions of the standard TEX fonts which probably was key to enabling them to display things quickly. Later on those fonts became available via CTAN which probably was a great step forward for the TEX community at the time.

I think people have tried to do simple minded TEXtures like things on today’s machines. Just auto-save documents periodically and let TEX run on them in batch mode. And I do wonder how far performance of that could be pushed by optimising things a litte, say, by storing intermediate statuses of TEX while processing the document and just resuming that processing before the earliest change in the file.

Anyway, TEXtures remained a dream for me but eventually I got DirectTEX which started bringing along its own MPW Shell clone at some stage. I switched to DirectTEX then and stayed with it. It featured simple project managing capabilities and also worked together with external editors (as did the other Mac TEX implementations). It was an all right replacement for the TEX Shell I used on my Atari in the end.

Just like the TEX system I used on the Atari, DirectTEX could also store all the temporary and junk files created during a TEX run in centralised folders. I quite liked that. And essentially I kept using DirectTEX until I switched to OS X.

A sidenote should be made on text editors here. In a way plain text editing is a very non-Macintosh thing to do. And thus the Mac wasn’t the greatest environment for text editors. After looking around a little, I settled for Alpha which wasn’t the prettiest beast around, but a rather powerful one. It brought editing modes for all sorts of languages and was particularly strong in its support for TEX and the various TEX applications you could use on the Mac. Back in that time it started to be clear that there are people who do ‘get’ (and are willing to fork out a fortune for) BBEdit and that there are others who don’t. I am with the latter. Unlike BBEdit, Alpha didn’t really make it to OS X, so

Next: Mac OS X

In my early TEX days I made two somewhat strange sidesteps. One was that I quite liked the application, let my friends know about it and even made our school magazine with it.

As a consequence my friend Jean wanted to use it as well. Alas, he only had an MS-DOS computer at the time. So I managed to get a copy of TEX for that system on a stack of 10 diskettes and there we went. Let’s just say there was an installer that ran and installed things on his fancy computer with even a built-in hard drive. But they Just Didn’t Work™ (™ quite possibly owned my Microsoft, and even if it isn’t I suppose it’s one of the few bits of vacuous ‘property’ they actually deserve) and it took endless hours of uninformed fiddling before we could get things to somehow do what they were supposed to do. Jean had no idea about computers beyong Prince of Persia. I had no idea about computers without menu bars and together we wer lost in this. And even when things worked, they seemed much more clumsy than on the Atari. (And his 9-pin printer sucked, too, hah!)

A much more fascinating experience was when it was my job during a school-time internship to help setting up a proper Unix based TEX system. Not only were the Sun workstation we used there wildly exciting and far beyond my scope of understanding, but the guy who coordinated this was a real Unix guru. The type of person who’ll dismiss a colour screen when he has a larger greyscale screen at hand which can display even more terminal windows at the same time.

I didn’t actually understand all the basic technicalities back then, but I think we downloaded the source and set up everything ourselves. Where ‘we’ doesn’t include me. At the time, a ‘fat internet connection’ was something like ISDN speed for a big network and just downloading all the sources took a day. Thanks to the magic of Unix that still worked! Once things had been compiled and put in the correct folders (or directories, as people without doubt said in that environment), I fiddled with the actual TEX stuff, setting up formats, additional styles, hyphenation patterns and all that jazz.

If memory serves, we didn’t just get plain and LATEX, but we also got niceties like (the probably forgotten, by now) Sli which people apparently wanted and appreciated. Quite an interesting experience for me at least. And it seems to have been useful for other as well. Back in the days where you couldn’t just download a ready-made pre-packaged installation for Unixy systems as you can today.

Next: Classic Mac

The first TEX I ever used was on an Atari ST. And that was pretty exciting, if only for the fact that it was the only application I knew of that was so huge that it required you to have a hard drive as you could easily fill ten megabytes or so with all the auxiliary files and particularly the font bitmaps. Gigantic!

Back then, hard drives came in large-ish boxes back then and had impressive cooling systems that – noise-wise at least – seemed to have been designed like jet engines. Certainly exciting things that were worth having at the time.

The good thing about my start with TEX was that I didn’t have to install it. Back then, installation could have been slightly tricky and, anyway, I had absolutely no idea about the inner workings of the thing, so getting a readily installed copy was probably essential.

It came with the well-known command line tools wich you couldn’t really use in the Atari’s graphical interface, a text editor (Tempus) and, the brilliant thing, the TEX-Shell. That shell tied all the components together nicely. And if I am not mistaken it was an absolutely essential tool for making things work because the Atari wasn’t exactly a multitasking environment and, for a bit of comfort, you needed someone to switch between TeX, the previewer, possibly Metafont, and the editor for you, which the shell did.

And, with hindsight, the TEX-Shell did that rather well. Not only did it switch between the various programs you needed to run, it also redirected TEX’s editing requests to the graphical editor, provided the ability to switch between different format files (even things that are simple packages today like AMS-TEX required special format files back then and there were also ‘Big’ ones which allocated more memory and didn’t run on my computer with just 1MB of total memory). And, most conveniently, it also provided project management: You could create a list of files belonging to your project, edit all of these and set one of them to be the ‘main’ file that will be typeset. Not mind-bogglingly fancy but getting the job done.

TEX on the Atari, or at least the installation I got, had another really neat feature which I have always struggled (and right now fail) to get back in later installations: It was set up to store all the output files in separate folders: a folder for DVI files, a folder for log files and a folder for auxiliary files I think. Which in turn meant that all that junk didn’t litter the folder you had your actual files in.

Other things I remember from back then: Speed. Or, rather, lack thereof. Just processing a single page document could take quite a while. If you went multicolumn almost a minute. Which made usage of the tools quite different from what it is today. Then there was Metafont. Also incredibly slow, but luckily not needed that frequently in regular usage. Except if you wanted to fool around and try typesetting musical notation, chess diagrams custom sized cminch characters or other fancy fonts — all of which I wanted of course.

And finally there was printing. On the glorious NEC P6, one of the more noisy machines on this planet. And one that, in its high quality mode, managed to need amazingly long periods of time to print a single page. Printing had to be co-ordinated with the parents’ (and neighbours’) wishes to sleep. Which essentially meant that you had 20 pages to print, you better started around eight, so no complaints would be coming when people wanted to sleep. The print quality was quite good though. In fact, I thought that most inkjet printers I saw in the late 1990s produced worse results when not being fed hideously expensive paper (i.e. most of the time). And you could do fun things with the printer such as removing the ribbon and making paper with fake watermarks, for example.

So, despite many things being quite archaic (TEX 2, slowness, no multitastking), it’s quite amazing how many things you could already do on that ancient machine. In particular, I keep thinking that the ‘Shell’ and the simple project management it offered did a rather good job compared with many other solutions.

Next: Sidesteps

The first thing to discover was that TEX could do all that maths stuff. While I guess I didn’t have particularly much real use for that back then, it certainly was amazing. Once you got your head around a few commands the computer would magically put lines, letters and numbers of different sizes together and create ‘exciting’ fomulæ. But there was more magic: running headers, footnotes, all sorts of numbering, tables of contents, dashes of different lengths, friggin’ ligatures! And if you hadn’t had enough with that you could even define your own commands.

In short, TEX totally rocked back then. Simply because it could do things that you hadn’t seen before. (And TEX was already a teenager by then). And, generally, I think TEX held up rather well over time. Sure, other applications may have reached some of TEX’s features over time: formula editors (which admittedly suck) in many word processors, support for technical writing in applications like Frame Maker or non-trivial line-breaking in applications like InDesign. And most of the others offer more appealing user interaction on the first sight, but if you can adjust to that, you’ll find that TEX holds up rather well.

Apart from the decent typesetting results, TEX is growing to arrive in the modern times in terms of the file formats it accepts for input and output. And with the more powerful computers we have these days, also limitations in the features TEX can provide in a single run seem to have fallen.

I think TEX is holding up rather well here. And that’s before we even touch all the technical niceties: Plain text file input that is extremely long lasting. Which other files from 15 years ago can you still use today? What’s the complexity of those files? How closely can you reproduce what you got back then? (Just compare TEX and HTML for a laugh’s sake…) And of course plain text offers all sorts of easy ways to automatically generating new files or using versioning systems on your files with the ability reasonably highlight the differences between versions.

Finally there’s the availability: TEX is as open as it can possibly be. You can get the source code and you can use it. Even better, it is code which is non-trivial and took quite a bit of research to come up with. Still it is available. And not just that. It’s even extensively documented and the author made a point of supporting it - even financially, by ‘paying’ people for bug reports - for a long time. Compare that to most of today’s ‘open source’ software which only rarely offers any of these benefits. Or to ‘closed source’ software which offers slightly different but not exactly more benefits albeit at a price. Apple paying you for finding bugs in their software? Microsoft making a serious effort that your Office files will be available to you in the future? Open source projects that aren’t byzantine? Laugh on…

While TEX is relatively constant, its surroundings aren’t. Running TEX posed different levels of a challenge over time and I will try to outline a part of how things changed and improved (or got worse) over time in the coming days.

Next: TEX on the Atari ST

And they are improving. Particularly when compared with other software. When I got into TEX in the early 1990s processing a document was an adventure. And when using ‘fancy’ packages styles, it could take about a minute to process a page on my Atari ST. And that was just processing the document, not even printing, which in high-quality mode on the NEC P6 took an eternity and kept the whole neighbourhood awake.

Now look at the progress: Today speed and memory are almost non-issues for TEX. You can load all the packages you want and it will happen almost instantaneously. Up to hundred pages per second aren’t unrealistic even on the smallest MacBook. And even the output of that, given as a convenient PDF file, has improved over time. I’m quite happy with that. (Bonus question: What were the speed improvements of XPress or Word in the same time frame?)

One thing that keeps irking me is TEX’s positioning of subscripts. Why the heck is the position of a subscript changed when a superscript  – and that includes the innocuous prime mark ′ – is added. It just makes things look silly when they are written next to each other.

I don't really know much about the 'semantic markup' business, but as far as my understanding goes, the main point is that your markup should attempt reflect your document's structure.

This is pretty obvious for things like headings – tell your browser it's dealing with a heading instead of telling it to use Zapfino at a certain size for those chunks of text. Easy. This situation seems pretty much win-win to me. I don't have any extra effort as I'd have to chuck in some markup for the heading anyway to specify the font. In fact, I may even save some work as the heading tags are particularly short. In addition my reader may enjoy the benefits of this markup if his or her browser can automatically generate an outline from the heading tags I used (not that I've ever even seen this rather trivial application of the markup in client applications).

So here, semantic markup – or whatever I think semantic markup is – works pretty well. To me that win-win aspect seems important. It isn't any extra work. You won't go insane by having to put zillions of extra tags in your documents. Just a different tag.

My impression is that advocates of semantic markup sometimes want to go too far: include markup for every abbreviation, state the significance of every single word by inserting extra tags. This usually means a lot of extra work for little to no benefit to the reader. So there's very little incentive to do it – in particular if the tags aren't generated automatically. If your objective is to write a text, you probably don't want to waste your time typing tags.

Let me digress to a topic that I am naturally interested in: mathematics and electronically writing about it in (La)TeX. TeX is frequently touted as an example for markup. I don't think (anymore) that's correct. Firstly because technically the examples aren't in plain TeX but rather in LaTeX. In LaTeX, to start a new section you simply type \section{Semantics} and there you go: a new paragraph, a heading, a bookmark in the PDF file, updated running headers, table of contents entry – whatever. If you don't like the looks of it you load a different document class and your heading will look differently. A bit like loading a different style sheet in HTML.

This kind of markup is nice. You can focus on what you're writing and don't have to worry about the looks. But what happens where the real work gets done? In formulæ? Consider the following:

xⁱ

I think that's a good thing. If people were seriously into marking everything up, they'd precisely state what each of the letters is and what the raising of the second letter means. Surely that would have certain benefits as changing a document to the notation that you personally favour would only cost a click of your mouse. On the other hand, however, document creation would probably grind to a halt as that approach would have an unacceptably low content-to-markup ratio.

An interesting question to ask in an all-marked-up world would be how that precious hobby of mathematicians known as abuse of notation is handled. Often people use notation that omits certain details. While those details may be important for comprehensive markup, their omission may actually make the expressions more human-readable and even comprehensible as the reader's attention is shifted to the significant parts rather than caught in notation.

Another question is how MathML fares. I haven't looked into it thoroughly. But my fear that it forces you to use all-markup all-the-time seems not to be true. There seems to be an option for presentational markup as well. Both of them look rather scary, though. Just check page 10 of this introduction for an example. It doesn't look like a winner to me. It tries to do too much.

And for good taste, let me show off what was forwarded to me from an unknown source:

And yesterday I bumped into another strange thing. This time concerning hyphenation of the word topology which was hyphenated topol-ogy.

I am aware that hyphenation is a strange area of the English language and I have heard many native speakers recommend simply not to hyphenate if in doubt. While this is fine for hand-written things, of course it is not an option when trying to do some justified typesetting.

From the information I have gathered so far, hyphenation in British English is done mainly along ethymological lines. So surely it should be hyphenated topo-logy [coming from what's probably spelled τοποσ and λογοσ if you like Greek letters]. In American English, syllables seem to be important for hyphenation. But again, with my sense of rhythm I'd end up with to-po-lo-gy, making T_EX's hyphenation look wrong.

While people seem to keep track of hyphenation problems in T_EX, I couldn't find the file of all words which are known to cause problems. In addition my reasonably detailed dictionary (Concise Oxford, 40000 words, not counting derivatives) doesn't provide information on hyphenation.

I guess the problem is that it looks like T_EX is wrong but I don't like to think of it being wrong, as it usually isn't.

Doing a Google search hints that on the internet the hyphenation given by T_EX is predominant, with more than 8000 hits for topol-ogy and about 450 for topo-logy. But then again, most – if not all – of these texts are computer-hyphenated and a sizeable fraction of them should be T_EXed, considering that the people who use the word topology and publish their writings on the internet are likely to be mathematicians or networks people.

Addendum: Will have to stop writing T_EX and revert to writing TeX instead. The spacing just looks horrible in web browsers and certainly doesn't match the subtlety of the original \TeX command. Also, the spacing of the Greek words looks horrible in Safari, but just fine in Chimera Camino.

Back when I bought that incredibly cool James Dean poster I had immediately thought that Jean will love it and got another copy for him as a christmas present. This way I had to decline his no presents this year offer and in turn got a record from him - the a.k.a. I-D-I-O-T EP by The Hives.