(That is supposed to be the Latin word ars (‘art’), by the way. Now that I write this I suppose that you could read it as if it were English, but that was not my intent.) Anyhow, correct me if I’m wrong, but I don’t think this particular combination is possible to write with any previous fonts, though it is in fact attested in Tolkien’s hand, in DTS 25. That specimen also exhibits an interesting lambe-lambe ligature. Hm…

However, the next thing I’m going to work on, probably, is a rather major change, but which takes place behind the scene: when finished, the resulting font will barely be noticably different from the current version. The change concerns the method to position the tehtar. Today, tehta positioning is achieved by means of anchor points: each glyph has a number of anchor points defined, and then the glyphs are attached to each other in such a way that the anchor points overlap:

While this is a fairly elegant and powerful technique, it also has its drawbacks once we need more precision in the positioning. It is difficult to place the anchor points in such a way that all the tehtar combine nicely with all the tengwar. In the example below, the tehtar and the tengwa óre all have their anchor points in the logical places, visually centered horizontally.

Now, where should the anchor point be placed on silme? We can align the i-tehta in a visually pleasing way, and then it will define silme’s anchor point. But, what happens if we then try to use that point to also attach the a-tehta? It won’t work very well: wider tehtar should ideally be adjusted to not collide with tengwar with ascenders. To deal with issues such as this, one could use more anchor points, introduce kerning after the positioning, use alternative glyphs, etc. Needless to say, things easily get very complicated.

That is one reason, but the main reason why I’m moving away from anchor points is that they are difficult to implement with AAT, Apple’s smart font technology, and I would very much want to have Tengwar Telcontar work on the Mac. So, instead of anchor points, I will rely solely on contextual glyph substitution: exactly as how tehtar are positioned in Dan Smith’s fonts, there will be a number of glyphs for each tehta, each with varying displacement, but unlike Dan Smith’s fonts, the correct glyph will be chosen automatically, and, for me as a designer, there is no specific restriction on how many different glyphs I may use — I foresee that I may use as many as ten or more for each tehta, to get exactly the positioning I want on each tengwa.

In short, I will have to rewrite a large part of the code, but the end user will probably not notice the difference. Unless they use a Mac, that is.

When I investigated possible ways to make my fonts, I quickly decided on Metafont. Not only is the concept of skeletons and drawing with pens an integral part of the Metafont language, I also found that this programming approach is especially well suited to the Tengwar script, because of all the regular and reappearing features. For example, I could write a function that produces a lúva (bowl), and then call this function from all the different tengwar. By editing this single function, the appearance of all tengwar is changed at once!

Donald Knuth chose the name Metafont to illustrate an important concept: the idea is that the metafont code should describe the abstract idea of how the glyphs are constructed (i.e. the “metafont”). In turn, how the actual drawing is made is controlled by a bunch of parameters. By changing a single parameter, you can at once, in a consistent manner, drastically change the appearance of all glyphs simultaneously, and thus generate many different actual fonts. A simple example could be to scale the glyphs horizontally. In the following picture, the first line demonstrates the effect of scaling the skeletons by the stated amounts. The pen then traces the skeletons after the scaling has been made. Note that the outcome of this is very different from the result you would get if you scaled the outlines of the standard font, as demonstrated in the second line. (Click on the picture for a PDF.)

Non-proportional scaling is usually regarded as a big typographical no-no, or at least something that should be done very carefully: beyond say 5%, the adverse effects of scaling on the outlines really begin to be noticeable. Scaling of the skeletons, on the other hand, is much more forgiving, in my opinion.

The regular and the bold version of Tengwar Telcontar are both generated from the same sources, but with different pens. Of course, there is nothing special with those two styles; they are both chosen from a two-dimensional continuum: the following picture shows examples of different pen proportions (again, click on the picture for a PDF version):

Other thing that can be modified in Tengwar Telcontar include the length of ascenders and descenders, and the x-height. In fact, there is no theoretical limit to how  many parameters a Metafont programmer can choose to implement.

So, if Metafont is so great, how come it is so rarely used? Well, there are a couple of important drawbacks:

• There exists no graphical editor. Metafont is a programming language; to define a glyph you will have to write code, and then compile it to see the graphical outcome. And if you’re not pleased, then go back and change the code, and recompile. Obviously, this design approach is not suitable for everyone.
• The output from Metafont is a bitmap font. To make a font in a contemporary vector format, this bitmap font must be converted, either manually, or with an automatic tracer (such as AutoTrace or Potrace). The automatic conversion is absolutely usable, but generally some manual cleaning up is needed.

For these reasons, I would not recommend Metafont to an aspiring font creator, unless they have a background as a programmer. But if you do feel attracted by the concept of “programming” a font, you might be interested in this tutorial. (I personally did not use it to learn Metafont, but from what I have seen, it seems like a good place to start.)

These Bézier curves are convenient to work with, and you can easily form virtually any shape with them. In fact, I found out when I started working on my first font that they were too versatile for my needs. So far, my fonts have been essentially calligraphic; that is to say, the glyphs are designed to look as if written with a pen. And when I attempted to achieve this by drawing the contours, I found that far too much time went into such things as ensuring consistent weight, forming nicely rounded line endings, etc. But if I were to write the letters using the method I wanted to imitate, i.e. with a pen on paper, I would not have to worry about things like that. All those features would come automatically, since they are defined primarily by the pen itself, independently from how I move it. (Obviously I’m talking about an “ideal” pen here.) What I wanted to spend my time on was to define the motion of the pen.

Thus, I decided to not use this method to make my fonts. In fact, with a few exceptions, I have not directly edited any of the contours of my glyphs. Instead, I made use of another concept, namely the skeleton, i. e., the path that the pen traces (shown in dark red in the following picture).

As can easily be seen, by defining the skeleton instead of the outline the complexity is considerably reduced: each stroke has two sides, but only one centre, so in that respect alone there is a halving of the number of Bézier curves that has to be defined. The outline shown in the first picture is made up of 23 nodes (although it could possibly be cut down slightly without disturbing the shape too much); in contrast, its skeleton is drawn with 10 nodes (not shown in the picture).

Of course, working with skeletons is much more limiting than working with the outlines. One way to get more freedom is to let the angle of the pen vary. For example, in Tengwar Annatar, the bar that closes the lúva in parmatéma and quessétema is drawn with a pen held at a slightly smaller angle to the baseline, which makes the bar thinner than it would otherwise be.

However, each individual stroke is drawn with a fixed angle, due to a limitation in the program I used to draw the glyphs. Since I was not entirely satisfied with this, for Tengwar Telcontar I implemented a way of varying the angle of the pen over the course of the stroke, which enabled me to achieve some subtle effects. As can be seen in the occasional grey pen marks in the picture above, the pen is held at a larger angle at the start of the lúva, closest to the telco, compared to at the end. Another, more illustrating example is the closing bar of parmatéma:

In the Tengwar Annatar at the left, the bar is drawn with a constant angle, which results in a fairly uniform stroke. In Telcontar on the other hand, the angle increases as we progress to the right, which makes for a more interesting shape.

My next post will be about how I have used METAFONT to draw my glyphs.

Lambe:

If I remember correctly, I had made an enlarged copy of Tolkien’s ring inscription (at an x-height of about 4 cm), and then I calqued selected parts of it, in an attempt to get an average lúva shape:

Note that I made these sketches primarily to simply get a clearer conception of what I wanted to achieve: I did not import them into a font editor and trace them there. Instead, the actual digitization was made largely from scratch, with at most some manual measurements in the sketches. More on this in a forthcoming post.

In my quest for a more original release, the closest I have come is this SIT-archive, which in addition to the same old TrueType font file also contains the illustrated documentation file and a Type1 version of the font, containing an extra glyph (the s-hook) – in other words, it is not a mere conversion of the TrueType file. The files in this archive are dated May 1993. Now, because SIT-archives are cumbersome to open, and because old Mac fonts need to be extracted to be usable on other operating systems, I have prepared an updated package of the font:

• Tengwar-Gandalf version 1

This version of the font is well known for its reversed lambe. The simple explanation for this somewhat curious shape is that it is a copy of the “T” in the public domain font Black Chancery. (In the same way, the rómen is a slightly modified “y”.) What is less known is that Mr. Elliott shortly after the original publication released an updated second version which features many improvements, such as refined u and o-curls, an extended character set including all the tengwar in the Appendix table, and a corrected lambe. This updated version can still be found to this date on various obscure sites and archives, but unfortunately it seems like it has never managed to reach a larger audience. (I for one had never heard about it.)

Now, more than 16 years after the original release, Mr. Elliott has generously provided me with the original archives, as published on a “BBS on CD” disc, and I have then extracted the font files and converted the illustrated documentation to PDF, to make it easier to read on modern computers:

• Tengwar-Gandalf version 2

There are still some characters missing compared to contemporary fonts, but version 2 is a huge improvement from the original and most common version, and it certainly deserves to be better known.

The tale could end here, but it turns out that there is yet another version of Tengwar-Gandalf. Supposedly around 1995 a certain Mr. Salamandir updated Tengwar-Gandalf version 1, in effect duplicating much of Elliott’s later additions, but also making further improvements not found in Tengwar-Gandalf v. 2. It can be noted that in the documentation for the font, Salamandir acknowledges that he had seen Tengwar-Gandalf, but claims that he worked from scratch when developing his font, despite the fact that many of the characters are obviously identical to those of Tengwar-Gandalf. Anyhow, this updated font, originally called Tengwar Medium, later Tengwar Sindarinwa, was published in 2002 and was available for a couple of years as downloadable shareware at Hybrid Elephant. (As from 2006, the font files are not possible to download, but offered for sale only; however, the old version of the site, as well as the font files, are stored at the Internet Archive. Update Feb. 2, 2010: The archived site has been blocked, at the request of Mr. Salamandir.)

The picture below shows the development of the two characters that originated in Black Chancery, lambe and rómen. The tengwar alda and arda were not included in the first version of Tengwar-Gandalf, and hence those in Tengwar-Gandalf v. 2 and Tengwar Sindarinwa were created independently from each other.

Then, in a bizarre turn of events, Salamandir is now accusing Elliott of plagiarizing his work. Tengwar Sindarinwa, he says, was originally designed in the 1970s and digitized in the mid ′80s (though not finished until 1995). Elliott must evidently have seen a preliminary version, and then copied it and released it as his own font. Or so Mr. Salamandir claims. For the full story, see the comments to one of my previous posts, as well as a blog post of his and its follow-up. From all that I have seen, he has no foundation at all for his allegations, which are inconsistent in several aspects and demonstrably false in others. Nevertheless, I want to thank Mr. Salamandir for bringing up this issue: If it weren’t for his attempt at muddying the waters, I would not have gotten in touch with Mr. Elliott, and would consequently not have found out about Tengwar-Gandalf version 2.

• The first a-tehta is displaced to the right. As you know, Dan Smith compatible fonts contain four different variants of the tehtar, so that the user can choose the one that fits best. In the original fonts, e.g. Tengwar Quenya, the lambe is not as wide, and the usual transcribers are adapted to that. In Tengwar Annatar (as well as in many other fonts) the widest tehta should be used here.
• While I am fond of the proportions of the lambe, it is way too big, and in a longer text it causes large holes in the texture of the page. If I remember correctly, I drew it this way so that it would combine nicely with the regular under-the-tengwa tehtar. I don’t know why I was not content with the special tehtar that are intended especially for lambe.
• Moving on to the next tehta, there are things to say about its position as well. Though not quite as noticeable on the wider tengwar, I would generally prefer to move the tehtar down a bit. Their high placement is a necessary adjustment to avoid collision with the nasalization bar.
• Ah, silme nuquerna. . . I fear we as a community have grown used to this shape, even though it is actually very uncharacteristic for Tolkien’s writings. As you might know, Dan Smith compatible fonts contain two variants of silme: one that is meant to have the tehta to the left of the ascender, and one that curves around the tehta placed to the right. Shapes similar to both of these are both common in Tolkien’s tengwar texts, e.g. in DTS 19/20. But when Tolkien draws a silme nuquerna, it is virtually always formed like a “9” — the shape seen in this picture is as far as I know not attested. Unfortunately it has by far become the one most commonly used, presumably because it is more natural to write on the keyboard and because it is the default in transcribers.
• This rómen is going to fall over; it is so unbalanced it is not even funny any more. Of course, it is an adaptation to avoid collisions with tehtar (and specifically to connect smoothly with the under-bar.)
• An example of the nasalization bar. It fits between the tengwa and the tehta, but it is rather cramped. Ideally, we should have some vertical kerning here.
• The tinco in itself looks fairly good actually, but the connection with the s-hook could be more smooth. There is a glyph specially suited for this in the alternative font, but in my experience, people very rarely use it.
• This s-hook is pretty awful I must say, with its quaint curvature. Thankfully, some of the other hooks in the font look better.
• Going over to the cursive/italic font, we find the same problem with the tehta on lambe, only worse. This is the kind of things people tattoo on their backs, and I can’t help feeling bad about it.
• I like the lambe, but there is a noticable gap between it and the following silme nuquerna. This is a place where kerning could be used, but it has to be done manually since it can’t be achieved with traditional dumb fonts, because the tehta on lambe intervenes. There is also an alternative shape available that connects to the following tengwa, and it should probably be used instead, to help bridge the gap. The shape used here is really meant for word endings.
• Silme nuquerna in itself is problematic in the italic font. In the ring inscription, esse nuquerna connects to the preceding tengwa, but there is also the need for a glyph when such a connection can’t be done smoothly, such as here. But I don’t think this glyph works very well; were I to design it today, I would probably draw it with a much smaller opening (cf. DTS 29–32.)
• The rómen . . . is fine actually. There is nothing I want to change.
• Here comes one of my pet peeves. Dan Smith provided two sets of nasalizers and consonant doublers: there are the straight bars, and there are the tildes. In the upright Tengwar Annatar, either can favourably be used (though I prefer the tilde), but there is no question that the straight bar is totally out of place in the italic font. Of course, the straight bar is the default in the transcriber I used, but the main fault lies with me, who included the straight glyphs to begin with.
• The same objection I had about the s-hook in the upright font also applies to the italic font.

To conclude, something like this would be better:

Ideally, many of the glyphs should be redrawn, but this shows the kind of improvements anyone can achieve by choosing alternative glyphs and applying some manual kerning. For a better result with less fiddling, some kind of smart rendering is needed. The Tengwar Telcontar font was initiated specifically to address these kinds of limitations imposed by the Dan Smith model — ground breaking though as it may have been in the mid ’90s, we can do better today.

Mach has also created a monospace font, FreeMonoTengwar, which, in the same spirit as Tengwar Telcontar, is intended to be a fully usable unicode font for Tengwar. Mach has quite ingeniously tried to solve the problem of tehta-alignment by means of ligatures, which have the not insignificant advantage of being supported by many more systems and programs than Graphite fonts.

Update: Except for XeTeX that is.

Just for the fun of it, I have attempted to count the number of Tengwar fonts in existence. I have listed vector fonts of any format and encoding model. (I regard TeX-fonts as vector fonts; however, low resolution bitmap fonts have been ignored.) Since I am here mostly interested in the number of unique designs, I have chosen to include also fonts that are not publicly available. The list is roughly cronological (at least in the beginning).

1. teng10 by Julian Bradfield
2. tengwar by Michael Urban
3. Moroma by Lawrence M. Schoen
4. Tengwar Gandalf by Michael S. Elliott. (Updated/plagiarized as Tengwar Sindarinwa.)
5. Gondolin by Michael Everson. (Used in the Unicode proposal, but not publicly available.)
6. Tengwar Quenya by Dan Smith
7. Tengwar 04 (03, 05) by Paulo Alberto Otto. (Also known as Tengwar Hereno.)
8. Tengwar Sindarin by Dan Smith
9. Tengwar Noldor by Dan Smith
10. Tengwar Naive by Johan Winge (me)
11. Tengwar Parmaite by Måns Björkman
12. Tengwanda Namarie by Deniart Systems
13. Tengwanda Gothic by Deniart Systems
14. Tengwar Cursive by Harri Perälä
15. Computer Modern Tengwar by Ivan A. Derzhanski
16. Computer Roman Tengwar by Ivan A. Derzhanski
17. Code2000 (2001, 2002) by James Kass
18. Tengwar Typewriter by Jeff Anderson
19. Elfetica by Ronald Kyrmse
20. Tengwar Elfica by Enrique Mombello
21. Tengwar Teleri by Josh Griffing
22. Tengwar Annatar Italic by me
23. Tengwar Annatar Upright by me
24. Tengwar Formal by Michal Nowakowski
25. Tengwar Mornedhel (Galvorn) by Josh Griffing
26. Tengwar Gothika by Enrique Mombello
27. Tengwar Eldamar by Måns Björkman
28. Tengwar Telcontar by me
29. Tengwar Beleriand by Enrique Mombello
30. Constructium by Kreative Software
31. The publications by Thorsten Renk use a unique font (partly based on Bradfield’s teng10).
32. FreeMonoTengwar by J. “Mach” Wust
33. Tengwar New (Light and Bold) by Marcin Przybys. (Not available for download.)
34. Elfic Caslon by Peter Wiegel
35. Greifswalder Tengwar by Peter Wiegel
36. Tengwar Optime by Peter Wiegel
37. Tengwar Tellepsalinnacontar by “Tencedil”

All in all, a pretty impressive number, I would say.

(This post was last edited on Aug 30, 2010.)