Text transcript of show #117. June 13, 2008

Transcription

1 Hanselminutes is a weekly audio talk show with noted web developer and technologist Scott Hanselman and hosted by Carl Franklin. Scott discusses utilities and tools, gives practical how-to advice, and discusses ASP.NET or Windows issues and workarounds. Text transcript of show #117 Sorting out Internationalization with Michael Kaplan Michael Kaplan is a Developer in the Windows International group and the author of the popular 'Sorting It Out' blog that is dedicated it all things '-ization.' That means Globalization, Internationalization, and Localization. This show is brought to you by the CYRILLIC CAPITAL LETTER A. (Transcription services provided by PWOP Productions) Our Sponsors Copyright PWOP Productions Inc. Page 1 of 11

2 Lawrence Ryan: From hanselminutes.com, it's Hanselminutes, a weekly discussion with web developer and technologist, Scott Hanselman, hosted by Carl Franklin. This is Lawrence Ryan, announcing show #117, recorded live Wednesday, June 4, Support for Hanselminutes is provided by Telerik RadControls, the most comprehensive suite of components for Windows Forms and ASP.NET web applications, online at and by the CodeBetter Blog Network, delivering tried and true solutions to real world problems for building better software, online at Support is also provided by.net Developers Journal, the world's leading.net developer magazine, online at In this episode, Scott talks internationalization with developer Michael Kaplan. Hi, this is Scott Hanselman and this is another episode of Hanselminutes, and I'm sitting here at TechEd with my friend Michael Kaplan, a developer in Windows International. How's it going, Michael? Thanks for sitting down with me. here. Oh, pretty good. Glad to be You know, I've noticed you for a long time and I think we ended up meeting because I commented on your blog with regards to my interest in Ethiopian languages, specifically Amharic. Actually, I think the first time it came up was that Turkish bug, the Ethiopic next. It wasn't much longer. Ah, there was a Turkish bug. Actually, what was interesting about the Turkish bug because I'm interested in internationalization and you're a full-time developer in internationalization, why is the Turkish language interesting for Microsoft developers? Unfortunately, it's most interesting because of a bug that commonly happens and that's got to do with the Turkish eye where they have a dotted uppercase version and a dot less lowercase version and they cased it the way you would expect where the dot less and the dot go together. We don't do that but it means that they expect something totally different..net framework supports it well but it means all kinds of apps get broken all the time because they case something they can't find the character later. I think that's your blogging. Right, Das Blog. It was actually broken at one point because of that, not all the thing, but part of it and then you tracked down the bug, you find the one line of code that is responsible for bringing everything down. Right, exactly and a very good example of that would be say that you had an application that was going to check for the word 'fail' in a textbox and the user typed in 'fail' f-a-i-l, all lowercase with an i with a dot on it and in your code you said textbox.2upper, you would get, in Turkish, F- A-capital I with a dot on it-l, and then you would check if that equaled 'fail' and it wouldn't, then your application would fail. Right and Makes total sense when you explain that the Turks have four eyes. But it was very not intuitive. But all the wonderful things in language about that all harmony and the way that the words flow together is lost on everybody. What people remember is that darn Turkish bug. Yeah, exactly. That is an example of you can't test your application on every single language. What are the languages someone should pick when they are trying to get a good coverage? That's definitely a good core language to look at because that's one of those main bugs. Romania came up recently as another one because of letters that they wanted added that don't exist in the code page but exist in Unicode, so it's a good way to test whether or not they're doing the right thing using Unicode properly, and Ethiopic which was another interesting case because there were three core languages that were scheduled to be added to Vista that I thought really strongly need to be because they've been Unicode for like five years and I was sick of waiting, and it was So this is an example where there is a language in the Unicode spec that you felt Vista needs this support. That Microsoft hasn't supported it yet, but they could have and it was a Singhalese, Mongolian and Ethiopic. Singhalese, we sort of got in, we sneaked in because the government wanted to help us. After Tsunami, they needed help actually collecting names and things so that we did an XP path so they could use Singhalese. So we knew that was going in because we have the font and everything. In Mongolian, they've done much work on. In Ethiopic, they also have much to work on. When you say Ethiopic, you're referring to what is called the fidel which is at the Amharic delivery. Page 2 of 11

3 Ethiopic I don't usually call it Ethiopic. My understanding is Ethiopic is a compromise thing that was done in the standard for the standard itself to come up with a neutral name for the script that isn't actually how even native speakers think of the script itself. And native speakers call it the fidel but there is the Latin script we call, like when we write English or French or whatever, Latin script because that's the root language. Ethiopian Amharic and the similar languages have their Latin which is called Geez. Geez is the -- you also uses that script. Okay. So rather than calling it Geez or calling it Amharic, you call it Ethiopic and that is the compromise. Okay, so one advantage when you think of Arabic with script and language where they are the same, people get confused all the time. So there is an advantage having a separate name although kind of a made up name does seem odd. Really. It would be like using Indian language, using the Devanagari scripts. Exactly. Someone who knows about Hindi and Gujarati and all these other languages wouldn't necessarily think Devanagari doesn't exactly roll off the tongue. Very true. It basically describes the script. So back to my original question. What are some languages that I would pick? Do you pick -- people usually say, well, English, Spanish and maybe Chinese. When I was doing internationalization, I would go Arabic, Chinese, and then I would make pseudo internationalized language that have funky accents and stuff like that. I probably wouldn't, I mean German is a common one people would use because it is a good example of a language that takes up between 30% and 50% more space in the dialogs. So it's a good way to check for dialogs that are too tightly spaced because you want English to look good but you don't want to make the localizer work too hard. You don't want clipped stuffs either. Ah right. That's a very interesting thing. You're pointing out that the length of the string change. pages were layout. The length of the string will Will change the way your English is fairly wordy too, but German is just more wordy and German is also a good one because it is a tier one language for many types of markets so people are using software but it's a good typical language to use. Then beyond that, getting the languages with other international issues, Turkish is a great one to use because of the very strict issues that happened with the Turkish eye. I usually pick Arabic rather than Hebrew. I like to one right to left and Arabic has more issues related to it because Hebrew is very simple other than being right to left. ligatures. Because it doesn't have any Exactly. There is no eye shaping and What is a ligature? Well, ligature is when you're combining two letters together and producing something that doesn't look exactly like they would separately. Like cursive script in English. Exactly. In fact, if you look at all the Uniscribe examples that are in the MSDN and in the platform of SDK, don't even mention Arabic. They give cursive Latin examples so that developers actually can say, "Oh, I know what that is," instead of saying "Arabic, I don't know it." Okay. It's actually a wonderful -- it's definitely what we're talking about. default. So Arabic is a cursive by Arabic is essentially, I mean, they are block script, it is actually a very cursive script. It's very flowing, it's actually very beautiful, not the kind of font you use every day but some of the fancier fonts look like the way topographers put it in, it's beautiful, but certainly you find more problems with Page 3 of 11

4 Arabic than you would with Hebrew because it is to your right to left, you have all those shaping things. I usually like to pick something like Thai as well, because Thai has other issues related to word breaking and you want those things to work well by default and sometime they do but not all components do so well, in fact, one of the problems that I have seen in.net framework is there are so many things that are, I mean, some controls are fully spaced out and some are just thin wrappers around something else so you end up with support that varies and you can find one particular word that doesn't work with the script where the other one will especially when you're using third party ones, and so the Thai becomes a good test of whether or not you can handle Uniscribe things like So what actually is Uniscribe? Oh, Uniscribe is the component in Windows that does all of the work to shape any script that requires any extra work. Interesting. By extra work, any kind of extra work, whether it is right to left or checking for illegal sequences, ligature formation So anything that is not just putting up straight English, Latin characters. Pretty much. When I was doing internationalization, we were working in banking, we thought about the German example and we thought about more complicated examples. One of the things with one of the bugs that we wanted to catch was pushing UTF-8 code all the way through the system and back and making sure that as it move from web server to middle tier, from middle tier to database and back, that we didn't screw something up. We didn't go turning into block squares, testing the system all the way through to make sure that something wasn't lost. Right and sometimes people skip the Thai case because Thai has a code page and they skip, they move right into one of the Indic languages like Hindi because then you can test. If anything ever gets into the code page and you're using Hindi, you d better hope they speak fluent question mark because that's all you have left and you re done there. Yeah. We worked with the Bank of Bangkok so we learned that really quickly. page whereas Thai does and it's another easy way to test for which is the question mark to know you're in trouble. So let's back up a little bit to some of the basics because we're both language geeks, we're both Unicode-type geeks, but I think we understand ASCII, and one of the demos I like to do when I talk to high school students because I go out to the command line, I type copy con foo.txt, type a, b, c control z and I go debug foo.txt and I go on look at it and we go, all right 67, 68, 69, and here is the character, so people understand some of the fundamentals of why is a -- because it is 67, I'm not sure, 69, I can never remember Oh no actually 41 and 68, I think has a decimal amount in Oh you think it has a decimal, right, because you are a full-time developer in the stuff. Okay, I'm thinking in base 10. This is going to be a huge problem, Michael. 65. It s 65, I guess I do those in my head now but that's -- I go to fewer parties because of that. Yeah, that is a problem. So then I go to like Yahoo China and I will select the Chinese character and I will paste in it in the notepad in the middle of my A, B, C and I will save it as Unicode with a Byte Order Mark and then I will go out and I will say Byte Order Mark and this is like that Byte Order Mark is just so, I mean I cannot tell you how many times I have received an where someone said, "I look in there with these fine two characters at the beginning of my string and suddenly everything looked bad." What is a Byte Order Mark, when do I need it, when do I not? Why is it sometimes there and why is it not sometimes there? All right, so the Byte Order Mark was actually put in by Unicode as a good way to indicate what the text is and UTF-16 is where the most common use of it was and UTF-32 I guess would be the most common. The idea being to mark what the text is and it's interesting because the character is FEFF. That's the character there, it also doubles as a zero-width no-break space and they decided it was a good thing to start with because there is no purpose to having zero-width no-break space at the beginning of the text so if you see it at the very beginning, then it must be a Byte Order Mark. For Thai actually, I've had clients who we have actually used Laotian. We basically used Laotian because it doesn't have a code You wouldn't see it because Page 4 of 11

5 You wouldn't see it anyway because it is invisible. In fact, there were many who have complained about the Byte Order Mark and I say I have never heard some much complains about something that you can't see, it has no width, yet megabytes are written about it. Certainly you've proven that you don't, that you're system is not understanding Unicode because if you ever see the thing, if it ever shows up with the question mark in some silly character, then you're obviously not just playing you re zero-width no I mean, to this day, most Microsoft internal developers actually use WinDiff to do their diffing and it doesn't understand UTF-8 so you always see that Byte Order Mark anytime that they have saved something in UTF-8. describes So it's a marker but also it The place where it is a problem for people outside of Microsoft, we sort of led to kind of our own people criticizing us we include the Byte Order Mark in UTF-8 as well where it is not considered necessary but sincere in optional things because you don't need to know the order of the test. You don't need to know whether the Because the Byte Order Mark display, it's not just a marker, it is actually describing an order of something. Exactly, but in UTF-8, the order is always the same. Its byte space but it's useful in UTF-8 because if you go to notepad and you have just asked the text and you save it as UTF-8 in the dialog, we don't have any metadata to remember that but we have this Byte Order Mark to put in but if you loaded up again and you didn't get what you've saved, people would complain so we said, well, let's just add the bad Byte Order Mark. Unfortunately it seems, and our biggest complaints still comes from people who write UNIX shell scripts in notepad which is a market that I don't think anybody ever imagine. In Windows, you know that writing in UNIX shell scripts. Exactly. Okay. This sounds like the most farfetched in the world but I have hundreds of s from people explaining how common this is so I can't even describe Really. I was thinking to myself when you were describing that Byte Order Marks belong in UTF-16, UTF-32, I was like huh, but I see them in all my UTF-8, and then you said why Microsoft put it down, I was thinking, oh no, what did Microsoft do now? But it sounds like it makes sense. It does make sense. It's just that it doesn't make sense from the standpoint of people who support UTF-8, not for the characters but just because it happens to work. Yeah. When you look at something like UNIX Shell Scripts, there is a specific protocol that you have to start with and it doesn't include starting with these other characters so it causes them problems. So most people spend their time in whatever their native language is and it just works and when things go wrong, it's when they switch over in certain localized things. Actually, I worked with a bank in Israel and they were not familiar with internationalization and this is such an ethno centric way of me to think and I was like, "How can they not be familiar with international, they are international." Well, of course they are not, they are domestic. You know, relative to Israel but it just worked for them. When you start inserting other languages, you make a truly globalized multi- language system and you mix things particularly when you're mixing right to left and left to right in the same exact text file. Things start to be really interesting. That's something we still do poorly, I think. I mean it's interesting. Hebrew itself has some elements that are left to right, basically numbers, and they certainly have some English text in. As a rule, those things work well. What doesn't work well is when you actually Hebrew texture kind of inserted in English like this example or vice versa. We're very poor with that like true bi-directional. Poor in what sense? Poor in the sense that you can't really ever hope to copy paste a sample and have it look the same wherever you paste it. You have to worry about the machine you re on affecting the underlying thing, I mean, there are markers you can put in that will strongly decide stuff that we never add. Yeah, I've had that happen before when doing samples where all grab like an Arabic headline, you know, successfully copy it and I'll paste it into some application. It is just not the most clever application and I'll look at it and I took care of it Page 5 of 11

6 in college but I've forgotten all but I know enough to say that doesn't look like a word right to left Okay, so I have sentences to be wrong. Yup and its often they seemed to left No, no, no, the system is right around. better here. to solve. And things have been flipped Our story could definitely be That's a very difficult problem Well, I mean, in practice it is actually easy because I go under a simple principle. I believe that a computer programmer should do all the work as well as an eight-year-old will be able to work. toddling yet. I don't think our computers are Well, they are not smart enough to do this yet but an eight-year-old who is looking at the text can tell you what order things should be in. So if I look at an MUI system for Windows MUI? Multilingual User Interface. And I look at the list of languages that are in the native scripts and I look at the fact that in Arabic, all of the scripts that have parenthesis in them on their list like you have think of one. Jews. Arabic of Morocco, Arabic of One that's actually on the list like Chinese simplified or Chinese traditional, the Arabic where basically all the parenthesis get flipped so they look wrong and I think even an 8-year-old looks at that and knows that it's wrong and can describe the problem and I think anytime the 8 or 10- year-old can describe it, the computer should be able to figure out. Well, that's the thing though, I mean, an eight-year-old can describe that if it feels wrong because the human brain is such a great pattern matcher that you can notice whether this ligature is wrong or that bit of kerning is wrong, but how do write an algorithm to figure it out? In this case, I mean, the simple rule is we basically put in, the parenthesis are considered neutral so they take whatever is surrounding them but if in the case where you have it s right to left overall The system is right to left. But in this case, you happen to have a bit of Chinese which is left to right in the middle of it, you want the left to right to control. You don't want the Arabic to intervene on any of the text in the middle. Okay. So you want to put a marker in front of that neutral character to say this is left to right and we have those characters, in fact if you call the get date format function and there is a flag they can put in that says insert those characters, so we do that except that we do not automatically and it seems like at that display level, it should be something to just notice the context and we're smart enough to write the algorithm for this, it's just that nobody really wants to jump in and do it. It's such a difficult thing, I mean the idea that we work for the 99% case but you never know when someone might be writing Arabic and wants to include the Chinese character and think about all the Cartesian products of different combinations of things that someone potentially wants to do. Someone is doing Tamil and Ethiopian sideby-side and it doesn't work. How many kinds of obscure bugs do you get when its only two languages that are combined in some strange way? Have you seen anything like that? Well, it's actually relative focused on specific languages that are kind of edge cases when you mix them in. So like when you include Arabic in the mix, you find a whole class of bugs because you are putting Arabic in the world. Sometimes I run the other UI languages. This is interesting. In XP, I remember, if I wanted to go and visit the Chinese website, I have to go into the regional settings under language tools and say install East Asian fonts, but under Vista, to be clear, everything is just there, right. They are there. So on any machine, even mom's Vista home, she can go and start typing Chinese immediately. Chinese. But she can't get the UI in Page 6 of 11

7 The UI requires, that's an extra thing you include. If you have Ultimate, you can actually install them directly. If you don't, then you can upgrade. That was definitely the decision that I didn't agree with. Which? Not being able to have a regular Vista home have Albanian a So you use this term MUI. Well, I use M-U-I, there's M-U-I? There's a weird, I don't know Maybe I'm the only MUI guy the... Okay that I agree with, but So we both agree that all versions of Vista should be able to have whatever language anyone wants to speak. Well, maybe that will change. Well, there were things that were interesting about it because most people I have talked to, they have no problem with, I mean, they are adding a language, they are doing something extra, they have no problem with paying a little extra for that but the ultimate, it seems weird to them because I want a home system, I still have two people in the family who speak two to four languages. My mom is one. So why do I have to buy ultimately in order to do that. Exactly, and people don't mind paying extra but they do mind the So there are two different things. There is having any version of Vista, I can buy French Vista or English Vista and I'm going to get either the English user interface or the French user interface. On both of those though, I can immediately start typing Chinese or even Ethiopian. So any Vista has got Ethiopian out of the box, you just go in and you enable it as the language and they can start typing Ethiopian in notepad. But if I want to have a Chinese start menu or an Albanian start menu, then I have to go and get this, is it MUI? and then you user interface, MUI It's either Enterprise or Ultimate And the term is multilingual No, no, no, MUI is actually what the teen who owns it calls it now but there's a second generation. So there is a difference between having an input language in the sense of being able to input this and maybe someone is using an English or French Vista and typing Chinese s to grandma versus running the entire user interface in that particular language. Yes, in fact, the list of languages that are fully translated turned into is smaller than the list of the ones that you can actually type stuff into. What are the numbers? 208 locales in Vista and I think there's over a hundred if you include the partially localized, the language interface packs there. Those are free. Okay, so what is the difference in the language interface pack and in MUI? Well, localization, one thing they know about it is its very expensive. Yeah, it's very. In fact, it's equally expensive to do it badly as to do it well. I have been doing localization work for years. My very first job at a college was to do English, Japanese, French on Windows 3.1 and one of the things I found in working with localization is no one will ever agree with the translation that the guy picked. Some people will agree but you'll never get complete assent for any of those things. I can't tell you how many times we'll spend all these money, we'll translate it into Page 7 of 11

8 Spanish, and I will show it to somebody and "Oh, we don't say that here." Actually Spanish is a fun politically charged case because most of the localization that happens in the United States is done from Mexican companies which the pure Casiano people and Spaniards don't like at all and they can tell the difference. I find traditionally, if you work with Quebecois French, the people in France don't like it but they can't always distinguish and so I think they don't always know the same way the Spanish can definitely tell because there are so many dialect and differences that you spot. All right. This is a great discussion. Let's take just a brief moment and thank our sponsors and we'll be right back. Do you know how to make the possible out of the impossible? Well, the.net ninjas at Telerik do. They just released a huge pack of web controls all built on top of ASP.NET AJAX that will help you build impossibly fast and interactive applications in no time at all. They have made the impossible possible in desktop development. If you think you can't have a carousel component in Win Forms, well, you can. Their Windows Forms Suite features a super powerful grid view control and 32 other crazy desktop components that will give you dazzling WPF-like features but in Win Forms. They do the same thing in reporting solutions with a design surface like nothing else. It looks just like graph paper, gives you advanced page layout capabilities, makes it feel more like a graphic design software than a reporting solution. Go check them out at telerik.com and be a.net ninja. Thanks for listening. This week's Hanselminutes is brought to you by CodeBetter.com. The CodeBetter.com Blog Network is made up of over 20 industry leaders and speakers who are passionate about delivering tried and true solutions to real world problems for building better software. These guys are not only our sponsor this week, but they are also my friends. The CodeBetter.com Blog Network, it's where industry leaders blog. You can find them at codebetter.com as well as Devlicio.us. So, I'm sitting here talking about internationalization and localization issues with Michael Kaplan from Windows International, and yeah, I've definitely found when doing translation work that people are very passionate about the words that you use and they are very frustrated when things don't either look or feel exactly right. You pick the wrong word and you can alienate an entire group. I understand that they have to change the time zone map inside? What was the deal with that? There was a huge Windows '95 scandal and this was when the lines where removed from the map and it just became these amorphous shapes of land masses. That was safer than trying to draw lines because none of it agrees with the lines. So people were disagreeing with the way of the time zone because it was outlining the edge of their country. that was in dispute. Exactly. Which may have been a border And when it is in dispute, you can't tell about places like Cashmere and things like that and not have somebody unhappy with the choice you have made. One of the thing that's hard about being Microsoft is there are languages that we can't support because of those things directly. Really, like what? Well, we can't do Cashmere, for example. You can't support the Cashmere language? Well, because you can't call it Cashmere, Cashmere because some people really feel like they own it, and you can't pick one of those places because the people who feel they own it disagree with each other about which parts they own. So what do you call it? Well, we can't call it anything because anything we call it would lead to somebody else being offended and people go to war over stuff like that and they can use as proof that they have some in the way of an argument. So does this mean that basically everyone who speaks Cashmere wants to run Windows just to lose, loses out? Well, they don't lose that, they lose out by being in the box, I mean, we don't have a font built in. I see. Can they get it? They can get a font. Oh I see. This is interesting because when I was doing Ethiopian work writing menus for local Ethiopian restaurants, Windows didn't support it so I would go find the font because some intrepid Ethiopians would go and build a font from scratch. Page 8 of 11

9 And we d figure out a way to like type it in and people would write custom IMEs and put method editors. So I assumed we could probably look around in the web and find if there is a language that Windows doesn't support, there may be a font though. Yes, and that's the other thing that becomes interesting when people talk about us supporting the language, they are often talking about they want everything there. They want to look at how, they want the sorting, they want the font, the whole package. We kind of think the same way where, and this actually gets back a long way back, but I was very concerned about Ethiopic, about Amharic specifically, because I really felt strongly we needed in, but there was some concern because we kept getting promised the input method by the vendor we're working with that we're going to have. So this is really an important note, that you've got your font and you scrolled out in your locale and your sorting but the input meta-editor, we've got all these keyboards that have mostly English characters on them, and you can certainly buy a Japanese keyboard and you can buy other keyboards, but Ethiopic and Amharic and Fidels got something like 236 characters, it's not an alphabet, it's still a syllabary... And they need an input method that actually will work intuitively for them and we were promise this that it would work, we could do whatever they claim, we could do with MSKLC, they won't have it any way What is that, MS? The keyboard layout creator, by which you just spell and enter keyboards. At that time I didn't really know to differ with them but it turns out then the vendor kind of disappears. There was some kind of a minor snafu. We were worried that we would have to pull out all of it because we can't say we support the locale and not what you type in it. I was morally and technologically in every sense opposed to taking it out because I said this has been one of those scripts that's been there for a long time. We have no business waiting. So, I pushed really hard on a different technology which is being used, because people found that creating IMEs was very complicated. Okay, so I wanted to make Scotlanguage or Klingon, if I want Klingon on support and I want an IME for Klingon, I would have to go and write a bunch of code. All right. I mean there are hundreds of interfaces, there are no good samples, it's very hard to do. I mean, people have done it but it's incredibly hard and because of that, in Vista they created a text-base system that would wrap the hard stuff and you create a text file with the configuration that's needed, everything would have happen automatically. Sounds cool. So I said, "You know, maybe we could use this and just build one and build something for Amharic." So we went through. So I worked out a technological issues which was easy because I know the language so I didn't have anything that would be useful for a native speaker yet but I said okay, so we can actually do this incline with the settings, so we started kind of trolling around looking for experts in the language who would be able to give us a big list that amounts to if I type this letter or these letters, this is what I want. Right and this is would not just be any old random Ethiopian guy in the street. This should be someone who knows the language. Well, we want a native speaker. I've written keyboards even for governments where they're transcribing information. The keyboard they would use is very different than what a native speaker would. Exactly. We want people who just whatever they intuitively inspect. In the end, it turned out that what we did was fairly intuitive because we've had external people who actually said that it was very quick for them to do most of the guesses were what they were expecting Right, and one of the things that's interesting about Ethiopian is that of course in English, you've got five vowels, in Ethiopian you've got seven and there s this is non-vowel. So and as a syllabary and when Ethiopians learned to say their alphabet, they don't sing ABC, like we do. They sing ha, hu, hi, hi, ho, ho, la, lu, li, li, lo, lo, and they go on this grid and it's these syllables, they go, the first column is la and it's the A's example, and then the second column is the lu Actually people who speak Japanese would be familiar with that kind of a grid. So, if you know katakana, then you would feel comfortable, or Hebrew, you would feel comfortable in Ethiopian because in Ethiopian is actually Afro-Semitic language, and when they type it, Page 9 of 11

10 if you're going to type PA, you would type HA, but there is actually two HA's. one is it. So, you have to know which So which HA did you mean because they will sound like HA, but the vowel is a different vowel. And it s subtle so then you have to pick, and we only have five vowels to pick on from the keyboard should you pick another one so there was a program that was called Ghez Pro on Windows 3.1 and that was the first custom input meta-editor and was basically a custom Ethiopian notepad. It only type Ethiopian and it only work with this one font and then you would basically type your Ethiopian script in this editor. So it's an all-in-one editor IME and then you would copy paste your work into word and Visual Basic 3.0 at that time did not support anything internationalized so people would create -- but interestingly it did at runtime, but the designer didn't. So what would happen is you would go and say, "I want a label that says first name." And you would write it in Amharic and Geez Pro, you put it in Clipboard, you paste it into the designer, it would show up as block squares, you would hit F5 and you run Visual Basic and then it would show up and look for What you wanted. That is still a way that many Ethiopian developers work today. Which is very scary. Which is really like that. But still some of that today, in fact, I was actually talking to -- I'm doing a panel later on SQL server and I was talking to the person I'm doing it with, they were asking me why it was that in the result's pane of when they are running queries, they were able to see any language but in the actual, when they are typing it, they just see square boxes. We kind of went through and I showed them what was going on, it's just some etic controls don't support peculiar languages, peculiar things but they were just floored at the -- some things just work perfectly, some things just don't work at all. Exactly. So in input metaeditor, they would type H and they would get a combo box like an auto complete with a top-down and it will show them here are the things that you might want. Exactly and as you continue to type, that list gets smaller So these are all lists of candidates. Basically and that is just literally what it is called, it's called the candidate list for IMEs and basically HA, and in our case, we have basically a lowercase A and H, uppercase A for the two different letters that were there and that's what we ended up choosing. The technology itself doesn't distinguish them. So basically it just shows the two in the same candidate list but maybe there would be some other letter that would be H, A, E or something like that so you actually maybe have more candidates as well. Eventually, you just get down to one or two and then you basically choose one and you're there. The typing is very fast although it would be faster if we put a difference between those two. It's interesting because to see more languages with more mature support like Japanese. You could use an IME, and you could type like the more got the one, you say TO and it pick a character, MO and it will pick a character, but for some people when the tablet PC came out, it was faster, they just draw it because they would just draw the stroke and that would provide the candidates so then there was that tablet PC based Japanese IME. Which people enjoyed a lot more and really all it requires is some work on the actual hearing recognition side which is easier for some language than others but also requires the effort and What was great about that and what's so interesting about it from a Japanese perspective is that stroke order is important. In the U.S. you can draw an A and it doesn't matter which side of the A you pick or if you draw the middle part first, there is no concept to that, but that stroke order can provide a way to narrow down your candidates. Although in English, I mean, I was told and I certainly did very poorly in penmanship going to school because for example Os you re supposed to start at the top of the circle Says who? Okay, says the teacher. Now I have always and to these days, still start from the right, that's just what I do. From the right of the circle? Page 10 of 11

11 From the right of the circle. I have no idea why, it's just what I do, but I get in trouble because you see the residual, you see some little turn of the thing is I always get marked off because of my Os that were in the wrong place. Apparently, 30 years later, you're still deeply bitter about this. I'm a little bit bitter about it. And I think that even though you failed on handwriting, that you're doing so well in internationalization. I think the best revenge is living well. Exactly. Well, thanks so much for sitting down with me today, Michael. I really appreciate it. Sure. Page 11 of 11