Here they are.

I’ve never encountered a romanization scheme for Thai (although it makes sense there is one). Do you have a link to it?

The sample setting you show is actually more Garamond than RdR, so if you want to match that, something as mundane as an Adobe font might do the trick for you. I would however recommend going the RdR route, if only to give the project a nice archaic look.

What are some digital RdR-style fonts?
If you don’t find a good one, can the project afford a commission for such a design?

The diacritics: well, the reverse-cedilla thing could be approximated with an ogonek (the Polish mark), which is available in many fonts (although with greatly varying quality), especially these days. As for the “horn” (which reminds me nicely of the Thai ngao), maybe take the visually comparable mark from Vietnamese? This will be harder to find, especially in an appropriate font, so you might have to graft one on yourself.

hhp

Thai has its own script (closely related to Lao & Burmese), which is not written with Latin characters. Modern Thai fonts are quite unrelated to this accented Latin stuff.

The “backwards cedilla” is an ogonek, byt the way. The little thing sprouting off the u is a horn. Offhand, I don’t know of any typefaces that would include these sorts of characters with the macrons as well, as precomposed characters.

I could get into the technical stuff in more depth, but the short version is that if you want to use a font in more places than Word 2003 for Windows, you’ll probably need specially encoded fonts (non-Unicode). I don’t know that such fonts with the characters you want already exist, but I can’t swear that they don’t, either.

Regards,

T

Thomas Phinney
Program Manager
Fonts & Core Technologies
Adobe Systems

Thanks for teaching me the word ogonek. I will be sure to use it soon!

Of course Thai has its own script but it also has a romanization system, which uses the special marks I described among others. We mainly need these to render titles in the bibliography, where I guess they are critical. We are using the system approved by the Library of Congress and the American Library Association and published in a document imaginatively titled ALA-LC Romanization Tables. Following up on Hrant’s suggestion I looked on-line for this and found it here: http://lcweb.loc.gov/catdir/cpso/romanization/thai.pdf.

I agree about the sample. I considered Adobe Garamond and may still go with it, since we already own that font, but was afraid it wouldn’t read as having the archaic quality that is being looked for. Also, this is only one of several model texts, and others are more RdR or Dutch in flavor than this one is.

Wouldn’t it be wonderful if we could afford to commission a special face! I doubt that would be within our budget, however (not having any real idea what it would cost).

Thanks for your help.

Commissioning a whole new face (even one based on a metal source) is indeed expensive — at the very least $2000 (for one weight). So try to find an existing digital face that has the right flavor, and just commission somebody to add the Thai stuff — or do that yourself!

BTW, it would be interesting to compare the LoC romanization to that of Lepsius (a scholar who once devised an entire standardized system of romanization usable for any script).

hhp

BTW, since most fonts include a free-standing macron, all you might have to do for that accent is add some big fat negative kerning pairs to the font (once you find the right font :-). And you could just create a free-standing horn and do the same for that — that way you’ll only take up one extra encoding slot, or even use unicode (which must have a horn somewhere).

hhp

Funny, I’ve been searching, and now I’m wondering:
Is it possible there are NO digital versions of any RdR-style face?!

BTW, I just realized: DTL Fleischman might work great here.

hhp

Interesting, I have never seen any standard for Thai romanisation before. Is trere any other Thai romanisation standard around?

BTW, does anyone have a reference to a history of Thai typography/printing? I just had an interesting in printing recently. But since I have been studying in Australia for the past five years, so I don’t have a chance to go back and search for information back at the Thai National Library.

Best,
Chanop

Finally, Typophile has a Thai member!  :-)
Hopefully Nathan Matteson (a Typophiler highly involved in Thai type) will join us shortly.

Since you’re here, please allow me to ask you something.
What do you think about the fonts in this sample?
http://www.themicrofoundry.com/other/thai.gif

BTW, I will dig up Lepsius’s Thai romanization scheme tomorrow.

hhp

Hrant,

I have been peeking around typophile for a long while, basically since I started having an interest in typography (was it typography?) when I was looking/learing how to make my thesis, report, and slide presentation look more interesting. I got bored with Knuth’s Computer modern, Times with MathTimes, and Lucida with LucidaNewMaths on a standard TeX system. There are not many choices there for mathematical oriented typography: Adobe just ditched a mac support for FrameMaker; hopefully, InDesign will have a built-in functionality to typeset mathematics and long documents to the level of previous FrameMaker.

For the font — I don’t have any training in type design at all — the first line looks nice; it probably is a bold/black weight of a text face.
The eyes of เ, ด, look optically imbalance, not sure why though.

In the second line, I have seen this type of typeface used often in a magazine. Some letters are quite far from the way we write though: ร that looks like s; and ฐ that looks like จ with a horn.

The main text looks pretty normal to me. IIRC, it looks similar to one that Microsoft includes in their OS.

I notice some spacing problems here — un-even spacing; I don’t know how to solve it from the font perspective without hacking into the algorithm of the text setting software though.

Regards,
Chanop

Fleischman is an interesting suggestion.

Regarding the Thai typeface sample, I will see if I can get some feedback from some of our Thai readers at the Asian Art Museum.

> I don’t have any training in type design at all

Like anything else, that can be both good and bad. Since there are many more designers than users on Typophile, and since the main point of typography is to accomodate the user, in this case I think that’s definitely good!  :-)

> In the second line, I have seen this type of typeface used often in a magazine.

What’s interesting -and maybe shocking- in this case is that this was taken from a daily newspaper.

> Some letters are quite far from the way we write

But what about their “distance” from what people expect to read?
1) Are any of the letters highly difficult to decipher?
2) Upon a quick initial glance, is there a problem seeing that line as Thai and not English (Latin)?

And a follow-up question: what do you think of the darker weight for emphasis in the body text? I mean instead of italics.

> I will see if I can get some feedback

Cool.

hhp

Oi vey. I was just on my way out to catch the first NHL play-off match. And here I am instead.

AFAIK, the Thai romanization scheme is used extensively in foreign language instruction and the like — but not used at all ‘in country’. By Thai that is.

>What do you think about the fonts in this sample?

[disclaimer: these comments are distilled through my Thai friends. I only wish they were wholly mine.]

Line 1: standard newspaper headline that wants the reader to think it’s more important than it is. Because of both the weight off the letters and the heads (loops). I.e. it’s ‘classical’ Thai. Much like I make fun of faces like Trajan, my Thai friends make fun of things like this.

Line 2: standard advertising, preferable to some readers than Line 2. Too romanized (e.g., 5th consonant, ro rua, is too much like an ‘s’.). A lot of Thai I know think the heads make reading harder and prefer faces as in Line 2. Even for running copy at small sizes.

Lines 3-4: The Latin is too heavy. Arabic numerals are too heavy and too tall. I wouldn’t be surpised if that was Browallia (an MS font as Chanop says). I’m no expert, but I think that the left sidebearing on sara e is too tight (7th ‘base’ glyph in the first word on the second line; 4th glyph in the last word). While it might look like the left bearings on sara o and sar ai are too big, that’s what I usually see. And what usually seems to be accepted.

Although, as I mentioned, all the head (loops) that make a face ‘look Thai’, they seem to hinder efficient reading. It’s something I’m stuggling with as I’m working now. To be headed or not be headed.

Sorry for the bad pun.

Thom: sorry for not properly addressing your post. I have a leaf from a 17th-18th c. book with a map of Thailand on the verso. I’ll scan both sides and post them this week if that would help. The map is copper engraved I believe. And I’m betting (I’d put money on it) that the text on the recto is set in something Garamond-esque.

Bloody hell. The Canadian games are on CBC anyway. Who wants to watch Nashville? Here are the pages if they’re helpful. 300 dpi TIFFs are available here and here.

Thom: I thought for a second that the romanization scheme for Thai might be similar to Vietnamese — and that you could use a Vietnamese font for this. But looking through my files, it doesn’t appear to be. I’m sure you’ve already looked into this though.

>But what about their “distance” from what people expect to read?

I’ve got two handpainted signs from a friend’s birthday party that I’ll have to post scans of Hrant. They’re in Thai — you’d love ‘em. I can’t make any sense of them whatsoever, but apparently Thai readers have no trouble at all. I imagine it’s their version of Bickham Script or something.

>And a follow-up question: what do you think of the darker weight for emphasis in the body text? I mean instead of italics.

‘Italics’ only appeared in Thai printing in 1925. That’s what the only source I’ve ever found says anyway.

OK, off to hockey finally.

Wow, the lc “g” on that antique map is like a Patpong exotic dancer.

Vietnamese: it’s not Romanized — that’s the away it is!  :-/
The Carmen Miranda of writing systems.

Nathan, I’d love to see those scans! But pre-emptively: there’s a difference between taking many seconds to read a strange display face versus reading a book.

> ‘Italics’ only appeared in Thai printing in 1925.

That’s good info. As a separate style, or subordinate to the “Roman”? And I wonder why a newspaper that would not mind that hyper-Latinized font would mind using italics.

hhp

Hrant,

For the font in line 2:
I have no problem reading it at all as it is on a head line. A quick glance, from my perpective, reveals that it definitely is Thai. I probably have some problem reading it in a long body text though as I feel that it is harder to read compare with the line 1.

For the body text, a number of thai font designers seem to settle down that bo-height (bo is the 10th character on the bottom line) should be somewhere between x-height and caps-height. In many fonts, the harmony between Latin alphabets and Thai alphabets are not there; quite often, latin counterpart has a thicker stroke; sometimes a different design.

The bold emphasis is a normal practice, I think. Usually there is no real italic (cursive) in same family but a slant form. To my knowledge, slant/italic shape in Thai came from a western influence.

Personally, I like Thai face with the loop. It is easier to read. If you have not seen this, the following two links are a publication that the goverment published a while ago about Thai fonts. They are in Thai though. The interesting bits for you guys probably are those specimen at the end of the book and the free thai font “Norasi” which is Yannis Haralambous’s design. The files are big (13.5Mb) and the difference between the two is Type 1 and TrueType font.

http://www.swpark.or.th/upload/download/ThaiNationalFont_PC.zip
http://www.swpark.or.th/upload/download/ThaiNationalFont_LINUX.zip

> it is harder to read compare with the line 1.

Why?
Is it totally because the structures are unconventional, or partly because your mind is gravitating towards the Latin “cognates”? Would you say that an unconventional font that’s not Latinized is less difficult to read (for a Thai who reads the Latin scipt)?

> bo-height (bo is the 10th character on the bottom line) should
> be somewhere between x-height and caps-height.

Makes perfect sense. This should apply to Cyrillic for example too, although I’ve found great resistance to the idea among designers of Cyrillic — which is particularly strange among their subset who also design Hebrew fonts, but for that script they do break the congruence!

In my sample however this doesn’t seem to be enough: “Home Depot” still looks too big. The balance seems to be more delicate than just “between x and cap”.

Slant/Italics: Why would you say Thai typographers avoid simple mechanical slanting?

Thanks for the links — I will definitely check them out!

hhp

Hey Nathan, Chanop’s PDF is the one you once sent me! Cool, less reading to do.  :->

hhp

Norasi: does it have a Latin counterpart?

hhp

Hrant,

It’s harder to read in a long text because of familiarity, perhaps. Most texts that I have grown up used a typical Thai face with the circular head. However, there is no problem reading unconventional face on a poster.

> Would you say that an unconventional font that’s not Latinized is less difficult to read (for a Thai who reads the Latin scipt)?

No, perhaps.

>”Home Depot” still looks too big. The balance seems to be more delicate than just “between x and cap”

Agree, totally.

> Slant/Italics: Why would you say Thai typographers avoid simple mechanical slanting?

I don’t think they avoid mechanical slanting; slant fonts are being used everywhere, however, I think bold are often used more frequenly.

Norasi came from an Omega-serif project — a free face that looks like Times Roman. The aim of the project is to support typesetting in many scripts e.g. latin, greek, hindi, hebrew, and etc.

Interestingly, Norasi used a scaled down Latin for its latin counterpart so that the whole font looks more in harmony. Nevertheless, the scaled down caps is ~71% while whe scaled down minuscule is ~77%. To my eyes, minuscule is still too big and too dark.

> bold are often used more frequenly.

Would you hazard a guess as to why?

> Norasi used a scaled down Latin for its latin counterpart
> so that the whole font looks more in harmony.

I don’t understand. Do you prehaps mean that the Thai font would look funny (too big on the body, I presume*) if it were made to match an existing Latin font outright? This would be an interesting paralles to my ArmenoLatin type system (Nour&Patria), where I discovered that the Armenian needed to be set about 10% larger to maintain apparent size (IF that’s something you want in a given layout). Unfortunately, this makes some people (mostly die-hard Modernists) very uncomfortable — and many of them seem to have a mental block towards understanding why this has to be so. What’s even harder is to explain to people why matching apparent size should not be a goal unto itself.

* But relative to what? I guess the bulk of existing Thai fonts?

hhp

There is a lot of good information here, and I haven’t had time to properly consider it yet because of the press of deadlines. But I am once again amazed at what a rich resource typophile is. Could this exchange have occurred pre-internet? Thanks to all of you.

Nathan, your samples are very interesting indeed. What is the title and what are the publication details of the book they are taken from? Do you actually have the book?

> Would you hazard a guess as to why?

Towards a modern look, or a contrasty look, perhaps.

> I don’t understand. …

“Too big on the body” I don’t quite understand the term, but I think it is. Let’s say we have a typical Latin face at 1000 emsq design size (Type 1); Caps hight is ~700, x-height ~450, and the length from the highest point to the lowest point (ascender-descender) is within 1000. To match this face with Thai alphabet that has a regular height between Caps and x height, let’s assume it to be 550-600. That is fine until we look at some characters like sara o, sara i, sara u, tonal marks, and etc. When we combine the total design size from the top to the bottom, it goes well beyond 1000, quite ofen 1300-1500.

Is there any real specification for this issue? Thomas? I have looked into Adobe Type 1 spec book, but found no concrete answer.

TrueType Font distributed with Microsoft OS tend to reduce the size of Latin alphabet; the exception probably is Tahoma which is designed for screen reading. Designer of Tahoma squished the upper/lower vowels, tonal marks a little. So everything would fit in the point size.

What I meant about Nosari is that, since the minuscule was scaled down less than the caps, so it looks too big. Besides, the stroke weight of Thai alphabets does not quite match the weight of the Latin ones, especially the bold weight.

Let’s see how I can post the image…. (still reinstalling panther, just got my pbook back from a service).

OK, so Thai has a “problem” with its distribution of vertical information: most of the letters are bo-height, so you have to make that big enough, but many of the letters use up much more vertical space. Sort of like accents in Latin caps (which are normally made to go beyondthe EM bounds), but I guess the super-tall letters in Thai are pretty common?

A request: could you possibly find me some linguistic data about Thai? Specifically, I’d love to have a chart showing the relative frequency of each letter in Thai. If you could, that would be very nice!

hhp

>Wow, the lc “g” on that antique map is like a Patpong exotic dancer.

I thought you might dig that ‘g’ ;-)

>Vietnamese: it’s not Romanized — that’s the away it is! :-/

What about Nom script? I’ve been under the impression that Vietnamese had it’s own ideographic writing system up until colonization and, consequently, romanization.

>Nathan, I’d love to see those scans! But pre-emptively: there’s a difference between taking many seconds to read a strange display face versus reading a book.

True enough. I’m still amazed at how divergent the letterforms can be.

>could you possibly find me some linguistic data about Thai?

Hrant, you might look up Doug Cooper at CRCL in Thailand. Also, the University of Wiscosin at Madison has a pretty good Southeast Asian studies department. I don’t know anyone there, but they might be helpful.

>[italics] As a separate style, or subordinate to the “Roman”?

Off the top off my head, I’d say a separate style. I.e., not subordinated to ‘roman’ in running copy.

>What is the title and what are the publication details of the book they are taken from? Do you actually have the book?

Unfortunately, I just have the one leaf. I’ve been half-heartedly trying to find some details about it, but so far without success. If I run into anything, I’ll definitely let you know.

It might be interesting to you that the first book printed with the Thai script was in 1828. About 60 years after the end of the Ayutthaya period. Before that every book used the romanization scheme you’ve got. Although these days when Thai is transliterated, it typically doesn’t use the macrons, horns, and such. That said, I think it’d be pretty hard to do ‘period-correct’ typography without building a new face with the requisite accented characters.

>TrueType Font distributed with Microsoft OS

You know, the Thai fonts distributed with Windows have actually gone through one huge change. Originally, Angsana (Times New Roman) was Angsana UPC. These fonts were made by Unity Progress, a Thai company, and I believe optimized for Thai — at 14 points the Latin was way to small. So now Angsana is Angsana New. As are all the Thai fonts in the OS. Although, perhaps the UPC versions are still used in Thailand.

Great thread. Sorry I found it so late. Chanop asked:

Is trere any other Thai romanisation standard around?

There is also an ISO standard for Thai romanisation, ISO/FDIS 11940, but I have not examined it because ISO want me to pay 79 Swiss francs for the privilege.

Regarding combining Latin and Thai in a single font, it is necessary to reduce the size of the Latin considerably if you want to balance it to the Thai and still allow adequate vertical space for the latter in the body height. Quite apart from the tall Thai glyphs, including three frequently occuring tall vowels, there are vowel and tone mark stacks to be accounted for, although the taller of these can reasonably extend beyond the body height. When working on a new Thai design, Fiona Ross, Tim Holloway and I examined a number of bilingual publications to get a sense of the conventional relative height and weight proportions of Latin and Thai when used together, and to see what works best. Sorry I can’t show the result yet; suffice to say that I can confirm Chanop’s observation that ‘bo-height … should be somewhere between x-height and caps-height’.

I had no idea Nathan was into Thai type. Cool.

> What about Nom script?

Yeah, Nom is not Roman, it’s Chinese-based. And I guess it’s possible there’s been a Romanization scheme for it, although Lepsius* came much later than the adoption of Romanized Vietnamese.

* BTW, the copies from Lepsius’s book that I’ve made don’t cover Thai. But here’s the reference in case anybody wants to look it up: “Standard alphabet for reducing unwritten languages and foreign graphic systems to a uniform orthography in European letters”, C R Lepsius, 1863.

> you might look up …

OK, cool.

—

John:
1) So do you have any letter-frequency data for Thai?
2) Check out Nathan’s Daasang on ultrafonts.com. (And while you’re at it, since I remember you having an ultra-hi-res display, please let me know if you think Mana-16 is usable for extended reading.)

hhp

Sorry, Hrant I don’t have letter-frequency data for Thai. I’ve yet to find an appropriately large and varied corpus of Unicode encoded Thai text, although the BBC archive would probably be a good start.


[Re. Mana-16, I find the variable contrast disturbing: the r is very dark compared to many of the letters with curves, such as g; the left side of the h is darker than the right side. I think I’ll stick with my ClearType thanks ]

> I’ve yet to find an appropriately large and varied corpus

In terms of word frequency a large corpus is important, but in terms of letter frequency a surprisingly small corpus will generally be very accurate. BTW, what do you use for “processing” the textual data?

Mana-16: I wonder if your gamma is unusually high. Anyway, Mana-13 will be coming out any day now, and it’s more “conservative” (mostly because the stems are 1-pixel). As for ClearType, if you like it, then just wait for handmade ClearType fonts!  :-)

hhp

Pat Chirapravati, who is associated with the museum and is contributing to the catalogue, says that both the more traditional top line and the more modern second line are pleasant, clear, and easy to read. Her only criticism is that some of the characters are too close together. She says they should never touch or come close to touching.

She also asks, how can we get a copy of the typeface?

> She says they should never touch or come close to touching.

Why not?

> how can we get a copy of the typeface?

Dunno — that was a scan from a newspaper — sorry.

hhp

> Why not? Just a cultural convention, as I understand. I don’t think it’s an issue of legibility, although it does seem consistent with the shapes, which seem to want to breathe.

> Just a cultural convention

So loose spacing is a cultural convention but traditional letter structures is not?

hhp

Tom, I think the top face is Quakerlady from UPC, and the second is Sathorn from DB. Where to go about finding them…I’ve no idea.

>some of the characters are too close together.

Do you know if she’s talking about sara e? If I’m not mistaken, sara e attaches to the consonant after it. In these examples it’s colliding with the consonant before — which doesn’t make much sense.

Hrant,

Here is a small study made by some friends a while ago. The result is cryptic in terms of encoding.

http://linux.thai.net/Members/poonlap/tmp/thaichar_freq.png

The graph is a histogram of the character number in hexdecimal in tis620 (iso8859-11) encoding, I beleive. The most frequent letter is sara aa which is the character D2. The raw data that he used is attatched below and you can find a sample tis620-encoding here:

http://thaigate.nii.ac.jp/refer/tis620.gif

If you need the data in another form e.g. glyph-frequency, please let me know.

Histogram of thai letters tis620_histogram.txt (1.1 k)

Wow, this is great! Thank you. And looking at the raw data, the corpus looks to be plenty huge.

A couple of questions:
1) What’s that complex character at DC?
2) For the floating accents (like in the range D4-DA and elsewhere), how does one predict how much above/below the main letter they will end up being placed? What base letters do they add on to? And can they double/triple up? What I’m trying to figure out is how much height/depth they can add to the vertical span of a Thai line (both typically and in extreme cases).

Again: thank you!

hhp

OK, looking a little bit closer, I can see that accents can double-up. Questions:
1) Which accents can double-up?
2) Can they triple up?
3) When accents double-up, are there rules of relative placement?
4) Would it be offensive to shorten the base character when an accent (or two) is added?

hhp

1. In Thai, letter can have a combination of a vowel mark and a tone mark.
2. Not for Thai, but I’m not sure if there might also be other, minority languages that use the script, and which might have different orthographic norms.
3. The placement of both is determined relative to the right-side vertical of the letter. However, in practical terms e.g. in OpenType GPOS positioning, the upper mark in a stack is positioned relative to the lower mark (<mkmk> feature), which in turn is positioned relative to the letter (<mark> feature).
4. Yes.

BTW, what do you use for “processing” the textual data?

The very excellent text editor UltraEdit has a count feature in its search function, which is handy for counting occurences of individual characters, pairs, combinations of base characters and marks, etc.

Hrant & John,

My friend told me that the data were collected from Thai web pages, mostly news, from a few years back.

>1) What’s that complex character at DC?

That’s the work-break (zero-space) character written in Thai, sorry. It’s quite important, at least for the moment, in technical implementation for line breaking and justification. Because a sentence in Thai is written without any space, therefore to break a like at an appropriate place, without breaking a word, is difficult — it still is an active research area. A few people/programs are using external helper who knows about thati words, to insert the word break with H&J engine so that Roman H&J engine can be used straight forwardly.

Only tone marks can double up, sometimes we have multiple slots for the same glyph (think Type 1), in order to accomodate the typograpical need i.e. top vowels have two variants: a normal position and a left-shifted position for letters like BB, BD, and BF; tonal marks have four variants, a normal position for combining with the top vowels , a lower position when combined with the normal consonants, and two left-shifted position.

Thai samples thaisample.pdf (7.0 k)

John: if I am going to process the data, I probably would write a little program to batch read a the text data; then feed it to a statistical package or process the data myself.

>So loose spacing is a cultural convention but traditional letter structures is not?

I am an advocate of tight letterspacing (for but most typefaces, as long as the word space is also tight,) but since when is it “traditional” for letters to actually crash, as some of the characters in these samples seem to do?

Moreover, as far as “tradition” goes, it is arrogant to apply Western traditions to non-Western languages. Traditionally, East Asian characters, for example, have very well defined spaces, far greater than in Western traditions. (Of course I realize there are “grass” styles that are roughly the equivalent of cursive scripts in Western writing.)

I am not expert on Thai language or culture, and so I must defer to those who are.

>Do you know if she’s talking about sara e?

Nathan, thanks for the lead on the typefaces. Pat is not around to ask anymore, but she pointed out the lack of space between the fifth and sixth characters in the top line (resembling “bi”) and the eighth and ninth characters in the second line (resembling “ak”).

Nathan, Tom: yes, that’s sara e on the first line who touches the previous consonant.

> The placement of both is determined relative to the right-side vertical of the letter.

Is it offensive to move the second (higher) mark to the left of the first one, to save vertical space? “Native” answers preferred.  :-) Non-natives can certainly have a lot of knowledge about a script (like John clearly does), but it’s much harder for them to have an implicit, “foundational” sensitivity. So Chanop, if you can answer this question (as well as my previous #4) that would be great. (Anybody else can answer too, of course!)

UltraEdit: I actually have it installed, but didn’t realize it can do all that!

—

Chanop, about the DC character: why is it missing from the Unicode Thai encoding? And why is its count so low in the stats?

> a normal position and a left-shifted position

So how offensive would it be to simply align the tone marks to the left whenever a (floating) vowel mark is already deployed?

—

> since when is it “traditional” for letters to actually crash
> it is arrogant to apply Western traditions to non-Western languages.

But then what about that second line?! I agree with your stance, and that’s exactly why I’m confused that a person wouldn’t mind hyper-latinized letterforms but would mind hypothetically latinized spacing…

Please ask this of Ms Chirapravati if you can.

hhp

> Is it offensive to move the second (higher) mark to the left of the first one?
> Would it be offensive to shorten the base character when an accent (or two) is added?
> So how offensive would it be to simply align the tone marks to the left whenever a (floating) vowel mark is already deployed?

Yes, those would probably be a totally new way of writing Thai script ;-) I think it’s orinigal, but Thai people probably have some problems reading it. One may think that it is a creative ads.

0xDC in iso8859-11 is 0x200B zero width space in the unicode chart; therefore it is not there in Thai specific one. The souce for the corpus came from html, so I guess those page were not using it at all or use a tag wbr instead. I could imagine some white noise in the data set as well.

Hrant, I see better now what you are getting at. I’ll see what I can find out (if anything).

Just as a conjecture: could cultural restrictions against inappropriate touching, which can sometimes be rather elaborate, influence the perception of type?

Along these lines, it might be interesting to see if there is elsewhere any correspondence between cultural attitudes to personal space and attitudes to spacing of type.

My guess is that any such analysis would be difficult or impossible because of the number of factors that come into play, and I can’t say that any obvious correlation in Western type jumps out at me.

For example, on average northern Europeans might demonstrate a greater ideal personal space than southern Europeans, but economic demands might nonetheless mandate tighter spacing.

Still, you would expect ideas about space to affect many kinds of cultural expression (architecture, for example) and consequently to affect typography as well. Somehow I am having trouble seeing this very clearly, however.

I suppose analphabetic languages would have a different set of rules.

> could cultural restrictions against inappropriate touching, which can
> sometimes be rather elaborate, influence the perception of type?

I think so, yes.
On the other hand, Arabic is the language of Islam!  :-)

I think a greater cultural factor in touching or not is a culture’s tendencies towards control, or Modernism. Westerners for example tend to like compartmentalized things which are easier to analyse, so they invented Greek writing with no ligatures.

hhp

> One may think that it is a creative ads.

Do you think it might be more OK in a bitmap font (for onscreen use)?

hhp

>be more OK in a bitmap font

I've been toying around with the vowels & diacritics in the bitmap font I'mn working on now, and my Thai readers say they prefer the standard placement and standard height.

Though it seems that mild vertical compression is far preferrable to horizontal shifting.

Though this a font meant for text. I'm sure with display work the constraints are totally different.

I'm sure they prefer the conventional scheme, the question is how much they would mind a modified scheme. Would they mind it enough to not use your font? And I think they would mind it less if clearly shown the advantages of economy (via lesser leading).

> mild vertical compression is far preferrable to horizontal shifting.

I agree. Ideally I'd like to do both though. :-)

BTW, making the base characters shorter when accented actually has a precedent in Latin: in some fonts accented caps are shorter. It's not common, but I think actual users mind it a lot less than other designers! Especially in a script such as Thai where the alignment zones aren't so strict, I think a laymen would have to have the discrepancy pointed out to him to notice it, and even then the typical response would be: "Huh."

Chanop, what do you think?

hhp

>how much they would mind a modified scheme

Well, they minded quite a bit reading the sample I gave them with vowels & such shifted about. So much so that I can't see any use for it in text settings. Not the particular approach I used, anyhow.

>if clearly shown the advantages of economy

But I think that you can only economize so much. E.g., we could modify Latin letters quite a bit (even more than we have) to make them more 'efficient', but I can't see people going for it. I don't see it being a worthwhile trade-off with legibility/readability. And I don't think that the average reader cares about economy nearly as much as the average typographer/type-designer.

>accented caps are shorter

Yeah, but typically those characters are infrequent (obviously barring all caps settings). In my experience, accented characters in Thai are much more prevalent, and I can't see that being an acceptable solution (to native readers encountering long amounts of copy).

The caveat, however, is that all of what's going on during reading is unconscious, I'm sure. All I have are conversations with Thai readers to form my opinions. I don't have any real (empirical) data to back any of this up. That comes in a few years hopefully.

Nathan: Uh. You're right.

--

Something else: I'm wondering, how are the "native" Thai numerals used? I noticed that Arabic and "native" numerals are mixed in that huge Thai font document, but only Arabic is used in the newspaper I have.

--

BTW, my brother is going to Bangkok for three weeks (in about a week from now). Should I ask him for anything? Chanop, what's a conservative newspaper I can ask for? The one I already have (please don't ask me for its name!* :-) seems to be on the "designy" side, with luscious color on every page, very high-quality thick white paper, and adventurous typography. I'd like to have an example of the other extreme (?) too.

* Although I can scan up the masthead if you like.

hhp

Chanop: Only tone marks can double up, sometimes we have multiple slots for the same glyph (think Type 1), in order to accomodate the typograpical need i.e. top vowels have two variants: a normal position and a left-shifted position for letters like BB, BD, and BF; tonal marks have four variants, a normal position for combining with the top vowels , a lower position when combined with the normal consonants, and two left-shifted position.

Tone marks used in stacks are traditionally made slightly smaller than those used directly above a letter. This both saves vertical space and ensures that the whole combination does not become top heavy.

Even when using something like the OpenType GPOS <mkmk> feature to build stacks, rather than having multiple non-spacing characters at different heights, a well made font will include these small tone mark variants that will be contextually substituted when preceded by a vowel mark.


Hrant: Is it offensive to move the second (higher) mark to the left of the first one, to save vertical space?

Questioning whether such a thing is offensive only addresses its acceptability to native readers, whose acceptance or rejection may be conventional. More interesting is to ask whether such an innovation is actually useful, and whether it makes sense within the structure of the writing system.

The answer to the question 'Is it actually useful?' has to be no. Apart from a small number of what may be termed double-width letters, e.g. tho phuthao and no nen, the Thai letters are not wide enough for diagonal left+down shifting of a stacked tone mark to reduce the total height: there simply isn't enough space in which to lower the mark. If you look at the first two pairs of letters in the example above, which are typical of the majority of Thai letters, it is clear that the vowel mark occupies most of the width of the letter. Even if one were to shift and lower the stacked tone mark on the wider letters, the total height required for stacked marks is necessarily determined by the narrowest base letter.

I also think your suggestion does not make sense within the organisational structure of the Thai writing system. If you look through the Font Book PDF that Chanop provided, you'll see that a lot of consideration is given to correct horizontal alignment of marks to letters (see pages 61-79). There is very clearly a correct way to position marks, which is relative to the right-side vertical stroke of the base letter.

The most common element in Thai text is a vertically organised grapheme consisting of a letter, a letter + vowel, a letter + tone, or letter + vowel + tone (some letters may also take modifier subscript marks). Because the 'sub-atomic' elements of the grapheme are arranged vertically, the verticality of the grapheme becomes an important structural concern. The verticality of the grapheme is established by the right-side vertical stroke of the letters, which is a feature common to all Thai letters: there are some letters that have no left-side vertical, e.g. ngo ngu and wo waen, but there are no letters that do not have a definite vertical stroke on the right side, even if some of these strokes do not ascend to the 'bo-height', e.g. ro rua, and in some letters elements, such as loops or flourishes, may extend horizontally beyond the right-side vertical, e.g. the loop in no nen in some styles and the tail of the ring in so rusi.

In order to maintain the verticality of the grapheme, marks above and below the base letter are aligned relative to the right-side vertical stroke. The only time they are shifted to the left is for the tall letters, e.g. fo fa, in which the right-side vertical itself rises above the bo-height. It is worth noting that when a tall vowel such as sara o follows a letter with a mark or mark stack above, it is the tall vowel that is letter-spaced (positive kerned) to avoid a collision with the mark(s): the marks are not shifted to the left to avoid the collision, because this would disrupt the verticality of the grapheme.

One might ask, why is the verticality of graphemes important? Why is it a strictly maintained aspect of the Thai script? I suspect it is a response to the Thai orthography and, in particular, to the length of visually unbroken strings of words making up phrases. Apart from light kerning to reduce white space caused by some combinations of letters, e.g. no nen followed by </i>cho chan</i> in styles where the lower right loop on the first letter extends beyond the right-side vertical, well-typeset Thai text consists of discreet, vertically arranged graphemic units that do not interact horizontally with their neighbours. This makes a lot of sense in terms of clarity when one looks at the long strings of such graphemes between phrase breaks. The verticality of the graphemes is an aid to breaking these long phrases into words, by preventing horizontal interaction across what may be word boundaries.

...about the DC character: why is it missing from the Unicode Thai encoding?

The zero-width space is not specific to Thai, so Unicode encodes only one, U+200B, which is used by various scripts and mapped to the appropriate decimal codepoint in 8-bit codepages.

> Questioning whether such a thing is offensive
> only addresses its acceptability to native
> readers, whose acceptance or rejection
> may be conventional.

Sure, and you allude to an interesting difference between the sensibilities of natives versus learners: the former are more demanding in terms of cultural authenticity (because they can feel that and they value it; and they can figure out "warped" glyphs), while the latter are more demanding in terms of explicit legibility (because they need that just for basic reading; and they're not good at discerning authenticity).

To make a font that's versatile but progressive, you would need to make sure learners can figure out what the new (or really "variant") letterform is, but maintain the necessary cultural sensitivity. If Thai people really don't have much of a problem with hyper-Latinized fonts, I think they might be able to handle some shifting accents too.

But to some extent you have to choose your target. And as with any culturally progressive effort, it makes more sense to target the natives and not the learners - for the latter one should just use the "normal" fonts.

Also, different scripts have different proportions of people learning (and needing to learn) versus natives simply reading. For a script like Thai (and unlike English) which might need to provide for its natives more than learners, a progressive approach makes more sense (as long as there's no strong threat of assimilation) than it would elsewhere.

> More interesting is to ask whether such an
> innovation is actually useful, and whether
> it makes sense within the structure of the
> writing system.

Well, of course that has to be established first - otherwise it's just change for its own sake, which I tend to see as a sort of hooliganism. The advantage of taming the vertical span of Thai is obvious: it allows for a better balance of apparent size and economy. Your issue of Thai letters "not being wide enough" I think is not only secondary, but possibly even a reason to explore improved readability through the introduction of greater width (especially variability of width) into the script! Or maybe you could vary the sizes of the accents (which you're doing anyway) - this size variation might even introduce more room to improve readability (in the way Hangul does).

There are many avenues to improving any script - you have to think from outside in, because that's where true functionality (as opposed to merely following potentially half-baked rules) comes from.

There's also another -probably more practical- advantage to taming the height: the need to reduce matching Latin (and other) fonts in size would be made less important.

Basically the way in which Thai is using the Cartesian space is somewhat inefficient: the extreme vertical spikes in a body that's mostly of uniform height throws a wrench in leading, hence economy. Imagine writing English where just the "h" goes up to double its height.

> There is very clearly a correct way to position marks

But the ever-present question in writing reform is exactly this: what is "correct", to whom? Is what's correct to a writing scholar correct to the layman reader? Not as a rule, no. Some people think there's only a certain way a Latin "r" can be correct, but when you corner them they have no real defense of their position. And you know what I think about them... :-)

People -especially as they get older- tend to like to apply what they know, instead of keeping an open mind. This is especially true in a craft, where there's constant pressure to simply and quickly produce more instances of something that works OK. But to some people "OK" is cultural death.

> Thai text consists of discreet, vertically arranged graphemic units that do not interact horizontally with their neighbours

Ah, but maybe this is bad? Sure, it makes individual "graphemes" easier to decipher, but like Latin text that's too loosely spaced, maybe it reduces readability? Always keep in mind the difference between deliberative and immersive reading.

hhp

John Tone marks used in stacks are traditionally made slightly smaller

Completely agree. BTW, I have not look much into OpenType spec yet -- sometime soonish, I think.

Hrant Something else: I'm wondering, how are the "native" Thai numerals used?

They can be used interchagably, however, at a less frequenty than Arabic Numerals. Also text in mathematics would not use thai numnerals in the maths part though; Latin & Greek alphabets and Arabic numerals are the norm.

For the newspaper, try Matichon, Thai Rhat, Daily News, The Nation. Sort of forgot, I read printed Thai newspapers and magazines only occasionally for the past five years.

Chanop, yet more requests, if you have the time and inclination:
1) I went through that entire 400-page (and enviable!) document on Thai type design, and I was wondering: could you please do a quick translation of pages 44-45 (with the table and the grid)?
2) What are the chances that the people who compiled that frequency data can run through the raw data again and find this: the frequency of "second-level" usage. I would presume there's a fixed set of pairs of characters (for example D4,E8) which would indicate usage of the "maximum height".

No pressure though, OK?

hhp

Hrant,

1) May be tomorrow as I am about to go out to the National Folk Festival in Canberra.

2) I will ask my friends for those raw data if they still have it as it is quite an interesting subject to me as well. However, this would probably take some period of time.

Cheers

Hrant, just to clarify: when I talk about Thai type I am expressing things that I have learned and ideas that I have developed in the process of working on Thai type design with experienced colleagues. I don't claim any great expertise in Thai, but my thoughts are grounded in the practical experience of making a typeface that has been described as 'extremely readable' by a fluent reader and expert in the Thai script. I mention this because, in your enthusiasm, you sometimes seems to assume that other people are beginning from the same level of ignorance or knowledge as you, and proceeding on the same theoretical and imaginatice basis. I think this can be a problem in our discussions, because I'm almost always commenting only from direct experience.

Also, your comments regarding the different sensibilities of natives versus learners are not at all in line with my experience in working with a number of scripts. Very often, I find I am much more concerned about the cultural authenticity of a design than native readers are, many of whom have grown up in a period of debasement of their national typography due to technical limitations, influence of outside fashions, or cultural uncertainty, and who may accept as normal things that are actually alien to or at best uncharacteristic of the longer traditions of their script. Unless they are themselves type designers, and working outside of contemporary fashions in graphic design, native readers are unlikely to do much research into such traditions, and so remain ignorant of what might be termed deep authenticity. I am not speaking here specifically about Thai; indeed, apart from some deplorable, latinising fashions in display type, Thai typography is in relatively good shape. But your generalisation regarding learners being unable to discern cultural authenticity does not seem to me to hold true. Cultural authenticity is something that can be apprehended from knowledge of a writing system and its development; natives and learners alike need to learn this, and can. This is not to deny that there are very insensitive non-native type designers making a hash of various scripts (there are also native designers doing the same thing), but they're just not doing a good job; they are, in fact, neither natives nor learners: they are not bothering to learn.

Hrant wrote:
Chanop, yet more requests, if you have the time and inclination:
1) I went through that entire 400-page (and enviable!) document on Thai type design...

Hi Chanop

I have been kindly lent the Thai Font book. (It has diagrams that are very similar to those I supplied to the Graphics Dept at Chula. Uni when I was teaching a short type design course there as visiting lecturer in mid-80s.)

Although the pdf of the Thai Font book is remarkably useful, would you know where I might purchase the book from?

Thanks
Fiona

John, of course I don't assume everybody knows as little as I do about Thai! But I do believe that different approaches and requirements allow for different types of "solutions". You generally make highly-usable fonts that "distill" conventional ideas into a highly polished -and not unoriginal- whole. That's very useful, no doubt about it. But then there's a parallel need to always doubt existing practice, to revert to the true nature of functionality and try to rebuild/redirect a script. Of course tradition is part of that functionality, but the lines you cross will depend very much on your intentions. For example, I think this is why you advocate breaking the Latin x-height bounds for Hebrew (which has varied conventions in that respect), but not for Cyrillic (because virtually no other type designers do it). Following the expectations of other type designers (and really not much anybody else, as far as I'm concerned) helps ensure greater adoption of your designs. If you made a Cyrillic font with an "x-height" greater than the one in the/a corresponding Latin font (clearly a much better solution for the Cyrillic user), the General would not be pleased, and the Bee would give you no honey! :-) But I do hope that soon will find yourself in a position to innovate more. Admittedly (assuming one has the patience) first establishing a reputation for solid conventional design might help more people eventually accept one's innovations, but many designers do seem to get stuck in a rut, never leaving the comfy (and materially more rewarding) practice of simply giving people what they're used to. I really really wish Slimbach for example would start leveraging his commanding position to push at least just a little bit, already.

As for your point about the types of sensibilities of different groups, I think you have to note how exceptional you are! Certainly, proportionally more non-native than native type designers have insufficient respect for a given script. I remember asking a few attendees of the Greek conference in 2002 what they thought of the current state of Greek type, and virtually all them disliked it as being inauthentic. Also, when you say that the average layman doesn't do any "research", you're right, but he doesn't need to! The point is that his sensibilities are built implicitly over time; he doesn't need to know why that second line in my sample is problematic to simply feel that it is. This is the difference between pleasing other designers versus the -more important- users.

hhp

> I remember asking a few attendees of the Greek conference in 2002 what they thought
> of the current state of Greek type, and virtually all them disliked it as being inauthentic.

{Add:}
While non-Greek type designers mostly don't see the issue at all. And this is normal.

hhp

For example, I think this is why you advocate breaking the Latin x-height bounds for Hebrew (which has varied conventions in that respect), but not for Cyrillic (because virtually no other type designers do it).

You are mischaracterising my reasons for harmonising features of the Cyrillic and Latin scripts. I have explained, at length, that within the specific sub-genres of post-Petrine text types in which I have worked to date, the Latin and Cyrillic scripts are historical and natural cognates. You can reject that style if you want, but the harmonisation with Latin is a feature of that style and has been for 250 years. If I were designing Cyrillic type in a different style, e.g. one of the Old Church Slavonic styles, I would not feel any need to harmonise x-height or any other feature to Latin, even if the font were to contain both scripts. Similarly, I can imagine new Cyrillic styles that would not naturally harmonise with Latin, but I'd rather leave these developments to the Slavs: I am in every sense a 'learner', and don't think it is my position to teach except when my work uncovers authentic but neglected traditions that contemporary practitioners are ignorant of. One might put it this way: you seemingly never met a writing system that you didn't think you couldn't improve; I never met a writing system that I didn't think could improve me.

By the way, when designing Hebrew type my choice about whether to align the head to the Latin x-height depends on the intended use of the font. I do prefer to make the Hebrew taller, and am happy that this convention is most common in Biblical scholarship, but when designing bi-scriptal fonts for modern office use in Israel I will align the heights, because this has become the prevailing convention in that setting.

Regarding Greek, I think you should ask non-Greek type designers who work with Greek what they think of the current state of Greek type design. I think you will find that most of them are very much aware of the issue of authenticity, and tend to be much more concerned about it than the average Greek graphic designer or typographer, who are the people responsible for the use of inauthentic, latinised types. I recall that your visit to the conference was fleeting, and that you missed much of the most interesting stuff. Were you there for the raucous open forum at the end, when Jean-Fran

PS. I think we should get back to Thai now.

I know you said we should get back to the Greek, but....

How would you characterize the Paratype folks on this spectrum? Is their Cyrillic more harmonized with the Latin than what you would do?

Just wondering.

T

> the Latin and Cyrillic scripts are historical and natural cognates.

I still don't get:
1) Why structural cognates can't simply be of different sizes.
2) How you can believe that a user benefits from such formalisms, which certainly comes at the expense of true visual harmony: mere geometric congruence between two scripts typically leads to discord in fact. This is obvious when you consider that different scripts use the Cartesian space differently.

BTW, how do you decide what two scripts are cognates or not?

> when designing bi-scriptal fonts for modern office use in Israel I will align the
> heights, because this has become the prevailing convention in that setting.

But you will be reducing de facto functionality by doing so.
This is what I mean when I say conventions shouldn't be given too much weight.

> I am in every sense a 'learner', and don't think it is my position to teach

Besides the exception you note, I can think of two ways where you should try to be a teacher:
1) By having the privilege of seeing writing systems as a whole perform in their own ways, you can gain insight into how each can individually be improved.
2) Don't underestimate the advantages of being an "outsider". Natives have some (probably most) of the advantage, but non-native can indeed be of benefit in providing a external, more objective view. When you're too close to something, you can't see it very clearly.

> you seemingly never met a writing system that you didn't think you couldn't
> improve; I never met a writing system that I didn't think could improve me.

You have also described me with your second sentence - certainly the two are not exclusive. Furthermore there are two aspects to what I'm saying, and you shouldn't bundle them: one is indeed script reform (like moving Thai accents around), while the other is clearly within the context of type design (like making the Cyrillic "x-height" bigger than the Latin).

Regarding Greek: yes, you have a point. But overall it's pretty obvious that non-natives generally care less than natives. This is common sense.

Thomas, you probably weren't asking me, but: I find that the Paratype Cyrillics are the same as most any other Cyrillics: way too harmonized on the superficial level, but as a result discordant on the true functional level - the bane of Modernism in fact. Anybody with an understanding of the difference between display and text type (as well as an ability to be objective) can see this clearly.

hhp

Oops, I meant Thai.

I've been working a lot on Greek lowercase in the last week, so it was just on my mind.

T

[More on Cyrillic, I'm afraid, that should be part of another, more general thread on type design across multiple scripts. I'd really like to get back to Thai before Chanop and the others lose interest in this thread: it's not like we get to talk about Thai type design very often.]

BTW, how do you decide what two scripts are cognates or not?

One looks at how they have developed. In the case of Cyrillic and Latin, these are have developed side-by-side since the 18th Century: this is the post-Petrine norm. There is a Latin letter a and a Cyrillic letter a: and in the post=Petrine styles they are the same form, weight and size. The fact is that, for better or for worse, the Petrine Civil Type was specifically designed to harmonise Cyrillic and Latin in this way: i.e. to make the same use of Cartesian space. This is the point at which I enter Cyrillic type design, with 250 years of dominance of this particular style, and with customers who want to pay me to develop useable, readable typefaces that conform to accepted norms for the script. Within this framework, I will make recommendations against prevailing norms if they are demonstrably contrary to the authentic development of the script within that culture. Quite apart from the longevity of the post-Petrine style, though, one has to see that 18th Century innovation in the context of a changing but still specifically Russian culture and its involvement with a larger European culture.

Hrant,

Here is the short translation

P.44
From table of letter relationship to ambiguous pair theory

The table of letter relationship use a simple crossword like rule: two adjacent
letters (top-bottom-right-left) must have similar shape.

Two benefits:
1 designing letters in group
2 magnifying some ambiguous pairs

Table of ambiguous pairs

ambigous pair | suggestion | note

1st row | the dip must be clearly seen | * and * are not amniguous because their heads are at different size

2nd row | make the tail intrude into the letter | when *, *, mixed with sara i, the tail might touch sara i; therefore making it difficult to notice the tail.

(mai hun-agad) (mai tho) | mai tho's head must show the diffrence to that of mai h
un-agad | At small point, this pair could be ambiguous especially in some unfamiliar specialised words

P. 45
Table of letter relationship

Fiona would you know where I might purchase the book from?

I have no idea; I asked my sisters to find some info for me sometime ago, but she could not find a place to by this book. In the front matter, it says that 1 600 copies were printed.

> the Petrine Civil Type was specifically designed to harmonise Cyrillic
> and Latin in this way: i.e. to make the same use of Cartesian space.

Even if I agree with your benchmark for "cognation"*, there's a problem with this particular statement: when I say "usage of the Cartesian space", I mean to include not just the shape and extend of structural elements, but also the frequency of usage of the different sections of the space. This is important because the visual information conveyed by text (and expected by the reader) is totally dependent on how often ascenders and descenders pop up for example - and of course Latin and Cyrillic are very different in this respect**. This is in fact why the problem with a congruent x-height comes up: because Cyrillic uses the ascender and descender space far less, it ends up having a smaller apparent size (same as in Thai, like Chanop pointed out). And in the case of a book in both languages for example, this will cause the Russian setting to look "less important" - which is clearly more of a problem than lack of geometric congruence at the mico level (something only designers notice anyway).

* Although even in the formal realm (not to mention the practical, "user" realm) it does seem problematic. For example, what script pair besides Latin/Cyrillic has cognates?!

** Heck, even Latinate languages -like English and Polish- are very different in this respect!

--

Chanop, thanks so much for the translation! I'll be looking at it closely tomorrow.

The book: If anybody can find a brick-and-mortar source for it, I could try asking my brother to buy a copy or two.

hhp

For example, what script pair besides Latin/Cyrillic has cognates?!

Latin, Greek and Cyrillic uppercase; again, in specific styles. I think it is a mistake to speak of scripts, per se, having cognates (in the sense of glyph cognates); rather, one should recognise that for historic and cultural reasons specific styles of writing and type have such cognates.

Not to sidetrack the discussion of Thai typefaces, but this is roughly the direction I think I am going with this book as a design project. Although the sample I showed was Garamond or something like it, there are also examples of early English printing that are in the general spirit we want to evoke, and I am tending more in that direction. This is black and white but the actual page will use color: the word SIAM maybe in a burgundy red and the fleurons in a very faint blue perhaps ... or possibly the whole thing will continue to evolve. ("SIAM" is a little faint here because it was in color.) I've never used fleurons like this before, and they're a bit in your face; my plan is to tint them back to where they're barely visible. Not sure the best treatment for the rules. I haven't fine-tuned the spacing of the type yet.
The page size is 8.75 in. x 12 in. The large outside margin might be used for sideheads.

Tom: I dig the border. Not often that you see that sort of thing. Off the top of my head, I wonder what it would look like if SIAM was in italic caps, tracked pretty loose.

>how are the "native" Thai numerals used?

Hrant, this isn't typographic, but I've been told that in school, Thais used to do arithmetic with Arabic numerals, but write the final sums with Thai. Not sure if it's still like that or not though.

Thanks, Nathan

We aren't using native Thai numerals.

I guess italic SIAM might look something like this.

> there are also examples of early English printing that are in the general spirit we want to evoke

My knowledge of Thai history isn't that deep, but I believe there were quite a few ties between the English and Thai. Originally because of the English conquest of Burma IIRC.

Just saying that an English typeface may well be as appropriate as a Garamond.

Hmmm. Maybe I was wrong about the italics :-/ Not sure I like that as much as I thought I would.

> Maybe I was wrong about the italics

It does add to the period flavor but I think it goes too far and somehow doesn't work as well as one would hope with those letters.

> I believe there were quite a few ties between the English and Thai.

Yes, I think that's the idea. Here is another example.

> early English printing

Then you sould really look at this:
http://www.hwcaslon.com/
Optical sizes and everything.

> the English conquest of Burma

In fact the French went after Thailand partly because they were jealous of the English imperialism in the region.

hhp

>it goes too far

Agreed. Do you have any samples of the interior pages, Tom?

Slightly off topic: There's a really good book about the colonial period of Southeast Asia by Thongchai Winichakul called Siam Mapped. If anyone's interested.

Regarding Thai numerals, Fiona reports that in many years of developing Thai fonts, including trips to Thailand, she hardly ever saw the Thai numerals being used.

Another Thai transcription scheme, at the top of:
http://ourworld.compuserve.com/homepages/profirst/h.htm#top

John, I've only carefully perused two Thai documents, but one of them (the type design manual thing) uses the native numerals more than the Arabic. The other (the newspaper), seems to avoid them completely. BTW, one great thing about the native numerals is that they don't lunge out at you like the Arabics (totally ruining the texture), but they still stand out a bit, aiding reading: even to a non-reader like me, their gentle perturbance of the texture is noticeable.

hhp

I've got a stack of Thai magazines at home, many of which use the Thai numerals. Although I concur with Hrant concerning the newpapers - hardly any of mine use them.

In country, I also noticed that most numerals were Arabic.

>but they still stand out a bit

Hrant, do you think this is due mostly to the fact they typically stand shorter than bo-height? Or because they have a somewhat different structure or 'character'? Or, of course, both.

> Thai magazines ... use the Thai numerals.

Good news! They're really much more suitable.

As for the reason they stand out, I think it must be mostly structural: they're round, while the letters are rectangular.

hhp

Hrant:
I've only carefully perused two Thai documents, but one of them (the type design manual thing) uses the native numerals more than the Arabic.

But not for page numbers or for listing point sizes

The other (the newspaper), seems to avoid them completely.

Yes, the fonts I was working on were primarily for newspapers and their sister publications, hence the almost total preference for Arabic numerals. However, we always designed and included Thai numerals for each font and weight.
(I understand that the numerals are akin to those used in the Cambodian script.)
I wonder if anyone has information about the reason for the alternative versions of the numeral 9.

Fiona

One thing to note about the Thai numerals is that their considerable variance in widths makes them unsuitable for tabular settings.

When designing a biscript font for Thai and Latin (using an existing Latin design but scaled and weight adjusted to harmonise with the Thai), I redrew the European figures* so that they were bo-height and blend better in Thai text.


* I don't like the term 'Arabic figures' because what this really refers to is the Arabic counting system adopted by the Europeans, not the eveolved forms of the European numerals, which are very different from their Arabic counterparts. The Arabs, meanwhile, call their numerals 'Indic figures', acknowledging the origin of their counting system, even though the Arab numerals bear little resemblance to any of the many forms of Indic numerals.

>unsuitable for tabular settings.

John, you might be interested in page 14 of the PDF Chanop posted. I'm not saying the numerals on theat page are the most beautiful - but I don't think tabular European figures are generally the prettiest either.

>don't like the term 'Arabic figures'

Point taken.

(I understand that the numerals are akin to those used in the Cambodian script.)

This is also my understanding. David Smyth writes, in Thai: an essential grammar, 'Thai script numerals are identical to those found in the Cambodian script, while the Lao script employs some but not all of the same number symbols.'

The Cambodian numerals were apparently adopted 'as is' from the Khmer (Cambodian) script. What I find interesting is that the Khmer numerals 7 and 8 are clearly the basis for the Thai tone mark mai tri and the vowel mai taikhu. I'd be interested to know why these forms were chosen for these marks.

John, you might be interested in page 14 of the PDF Chanop posted. I'm not saying the numerals on theat page are the most beautiful...

They're actually very impressive. Having had a heck of a time making nice proportional Thai numerals, I have a lot of respect for the person who made these tabular forms.

> unsuitable for tabular settings.

Really? Yes, they seem to require some gentle distortion to make fixed-width, but on the flip side there's no member that's totally hopeless: think of the Arabic* one. The worst of the Thai numerals (in terms of making fixed-width) seem to be on the order of the Arabic "4" and "7".

* I'd have to think about the issue of there being enough of a problem with "Arabic" to justify a break with convention. BTW, the European Arabic numerals look different than the contemporary Arab numerals because the Arabs stopped using the ones they passed along to Europe in favor of a set from India.

hhp

BTW, there doesn't seem to be a lot of handy info about the numeral "evolution", but I have found this: "The History of Numerals" by Henri Friedlaender, in Herbert Spencer's "Typographica" magazine, New Series, issue #1, June of 1960. It includes some very interesting samples of incunabula numerals.

hhp

Hrant: 2) What are the chances that the people who compiled that frequency data can run through the raw data again and find this: the frequency of "second-level" usage. I would presume there's a fixed set of pairs of characters (for example D4,E8) which would indicate usage of the "maximum height".

Just took a little break from study to write the code. Here is the character frequency data: single character, two-character sequence, and three-character sequence.

Thai character frequency frequency.pdf (679.0 k)

The corpus is Thai web pages, approximately 660MB in size. The data in pdf file are quite raw; I only check for thai letters and didn't specifically check for the second level usage that you asked. However, look through the two/three-character sequence. The info should be there. It should show some awkward combinations, hopefully.

Chanop

Wonderful! Thank you so much for this, Chanop.

I've converted Chanop's frequency data into Unicode encoding, and have re-sorted according to frequency, as this is most useful for my immediate needs. Here is a text file for anyone else who wants to play.

Thai characters and sequences by frequency Thai-by-frequency.zip (368.0 k)

It would be great if there were some easy way to filter out the spelling errors in the low frequency sequences.

Wow, Chanop, that's really great of you! We need more people like you. In fact it's hard to find data this good for English.

And John's sorting is a godsend - I got queasy looking at the raw data... Just one thing though: I don't have a Thai font installed, and would prefer keeping my system lean - could you please make a PDF out of that?

> filter out the spelling errors

Why? Spelling errors are part of the data too - some could even be intentional*. The way I see it, if it gets set, we have to worry about it.

* Like the way Cigar Aficionado gets their web site to come up no matter how you [mis]type it!

hhp

Hrant, here is the sorted one:

Sorted character frequency frequency2.pdf (732.3 k)

, same data.

John, filtering low frequency out e.g. percentile 95 and over is easy -- a one liner fix in the code.

This is golden. I owe you a Singha. Or three.

But there you can see the bad news: the fifth most frequent pair is a top tier riser...

hhp

No surprise, as the combination comes second in the three-letter sequence; "Tee" is an often used preposition -- one meaning is "at".