The History of Hangul Character Encoding

Where this story gets tangled most often is not in the dates, but in the meaning of the numbers. Whenever someone asks, "How many characters are in Hangul?", we first have to clarify what is being counted. Because there are four entirely different answers.

First is the number of jamo (letters). The individual letters taught in schools today number 24, while Hunminjeongeum originally created 28. Today, the international character encoding system sets aside 357 jamo for combining. The gap widens this much simply because each counts a slightly different set of things.

Second is the number of combinable modern syllables. There are 19 initial consonants, 21 medial vowels, and 28 final consonants—including the case where there is no final consonant. Multiply them, and you get 11,172. From ga (가) to hih (힣), that is how many syllable blocks can be formed under the rules, regardless of whether they have meaning. This even includes blocks you might use once in a lifetime, like bwaelk (뷁).

Third is the number of syllables actually included in a standard. This number varies from standard to standard. One standard included 2,350 characters, while another included all 11,172. Fourth is the number of syllables people actually use often. The syllables that appear in a single newspaper issue or a single document are far fewer than one might think. The gap between the third and fourth numbers is the true stage for this story. This is where judgment calls split: between believing that covering frequently used syllables was enough, and insisting that no unwriteable character should ever be left behind.

Hangul was designed by creating individual letters and assembling them to write syllables. The Hunminjeongeum Haerye (Explanations and Examples of the Correct Sounds for the Instruction of the People) includes an entire dedicated section instructing that initial, medial, and final letters should be combined from left to right and top to bottom to form syllable blocks. Yet computers could interpret this design in two different ways: by assigning a number to each finished syllable block, or by assigning numbers to individual jamo and combining them. Each method had its own strengths and weaknesses. And Korea was right on the verge of deciding which of the two would be enshrined in the core text of its national standard.

First, there was the method of assigning a number to each syllable. You chose which syllables to use in advance, made a list, and numbered them in order: one number for '가', the next for '각'. This is commonly called the pre-composed method, or Wanseong-hyeong.

This approach had four distinct advantages. First, the two bytes fit seamlessly into the slot layout used by international telecommunications protocols of the time. Second, in that layout, both bytes fell outside the range of ASCII English characters, so older programs scanning bytes for English letters wouldn't accidentally tamper with the inside of a Korean character. Third, if you arranged the list in Ganada (alphabetical) order, the code values naturally followed that order, making sorting effortless. Fourth, Chinese characters (Hanja) could comfortably coexist within the same code set. There was only one downside: if a syllable wasn't on the list, it had no number. There were a few workarounds, but we'll look at those later.

The other approach was the method of assigning a number to each individual letter (jamo). Instead of numbering whole syllables, you assigned a number to each initial consonant, medial vowel, and final consonant, combining the three to produce the syllable's code value. Commonly called the combinational method, or Johab-hyeong, this design split sixteen bits into a leading single bit and three chunks of five bits each. It had three advantages: all theoretically possible modern syllables could be expressed, individual letters could easily be extracted for search and linguistic processing, and adding more letters left room to expand into archaic Korean characters.

Yet it also had a downside. Encoded this way, certain syllables ended up with one of their two bytes matching the value of an English character. For example, the second byte of '게' shared the exact same value as the capital letter A. A program scanning through a file for 'A' would mistakenly grab the inside of '게'. Out of the 11,172 modern syllables that can be formed, 3,724 carried such bytes.

And here lies the central dilemma of our story. At the time, the slot space that international telecommunications protocols reserved for a two-byte character set was 94 times 94—meaning exactly 8,836 slots. There were 8,836 slots, but 11,172 modern syllables. There were simply more characters than slots. To abide by that standard layout, you had to pick and choose syllables; to include every syllable, you had to break the layout. Before even asking which approach was right, within the confines of a single two-byte system, it was a problem with no answer.

We need to be careful with the word "first" here. Depending on what is considered the first, the answer splits in five different directions.

The first national standard containing individual Hangul letters was the 1974 edition. The first national standard to represent syllables through combinations of letters was the 1982 edition. In other words, the combination model (Johab) was indeed an official government standard for a time. The common claim that "Johab was never a standard" is mistaken on this point. The first national standard to assign a distinct number to each syllable was the 1987 edition. What happened in 1987 was not the arrival of a new standard, but a replacement of the standard. The remaining two milestones appear later: the first national standard to include all modern syllables was the 1992 edition, and the first time the global character set included all modern syllables was Version 2.0 in 1996.

In March 1987, the government established an 8-bit, 2-byte completion-type (Wanseong) Hangul and Hanja character set as the new standard. This decision was based on research led by the Korea Standards Research Institute for just over a year. Looking at the character allocation table, there were 2,350 Hangul characters, 4,888 Hanja characters, and special symbols, individual letters, and Roman characters, totaling 8,224 characters. In addition, 95 slots were set aside for users to define Hangul characters themselves. This standard was soon adopted as the official code for the National Administrative Computer Network.

Why was this option chosen? The reasoning of those who set the standard was just as we saw in the previous chapter. It aligned with the byte layout of international telecommunication protocols, prevented bytes from colliding with English ASCII characters, ordered code values in alphabetical Ganada order, and could accommodate Hanja alongside Hangul. There was one more factor: the judgment that all characters frequently used in everyday work were included. Later, defenders of the standard responded: "With 2,350 characters, 99% of our Korean language can be expressed, and for that unusable 1%, utilizing the 95-character user-defined area is sufficient."

Yet the arguments of the camp that was pushed aside still remained: every modern syllable could be expressed, syllables could be broken back down into individual letters, and there was room to expand into archaic Korean letters. The two sides weighed different risks heavily. One prioritized the risk of breaking compatibility, while the other prioritized the risk of leaving behind characters that could not be written.

Lists began to emerge almost immediately after the standard was finalized. Korean linguists pointed out 220 characters that were missing despite being necessary for Korean orthography, and seven more were added later. The list was divided into clear categories: single-morpheme spellings, Sino-Korean readings, abbreviations, contractions, dialects, and phonetic transcriptions.

The roadblocks were concrete. When teaching spelling, teachers couldn't display a common misspelling of *kkeokda* (꺾다) on screen to show what was incorrect. When teaching standard pronunciation, they couldn't write out how *ot* (옷) actually sounds, because that character simply didn't exist in the set. Many characters found in dialect glossaries were missing too. Being left out of the standard was not just a problem for academic research labs. Like Korean language education and dictionary compilation, the very places dealing with standard language norms were the first to get stuck.

There was, in fact, one solution prepared: 95 slots left for users to fill in themselves. But there was a paradox here. If one person assigned *nuing* (닁) to the first slot and another person put a different character in the exact same spot, their files couldn't read each other's text. The value of an encoding system comes from everyone using the same table, but telling each person to fill it in themselves undermines that very value. That very criticism was raised even back then.

There was another detour built right into the text of the standard itself. It involved placing a Hangul filler character first and then stringing together the codes for the initial consonant, vowel, and final consonant. Written this way, a single syllable became eight bytes: two bytes for the filler character, and two bytes each for the three jamo letters. Here is the fascinating thing: to write characters outside the list, you ultimately had to lay out individual jamo. The principle of combination had already entered the precomposed standard as an emergency exit. However, because it was eight times as long and very few programs knew how to handle it, this method was rarely used in practice.

There is a scene that circulates most widely in this story: the tale that a TV drama broadcast in 1990 had a character in its title that fell outside the standard, so people could not even type out the title. Later newspaper articles cemented this narrative by pairing it with an anecdote about a beverage brand name. Let us separate what can be verified. We know for certain that the drama aired from April 16 to June 12, 1990, and that the character in question fell outside the national standard's primary character set. What we could not find, however, was any contemporary record showing what actually occurred within the computer systems of broadcasting stations or newspaper newsrooms.

What matters more comes next: what the original criticism was. Between 1987 and 1989, the arguments raised by the academic community were not about television titles. They focused on orthography education, standard pronunciation transcription, dialect records, and dictionary compilation. And there was one more argument, largely forgotten today: the warning that if each syllable is assigned a whole number as an indivisible block, a separate program is needed later to break them back down into individual letters. That critique looked ahead to text recognition and machine translation. A single widely repeated phrase captures far narrower ground than where the actual debate stood.

Furthermore, the sentence "Pepsiman riding a jeep and short-legged Ttombang Gakha" (찦차를 타고 온 펲시맨과 쑛다리 똠방각하) was never a factual record of an incident. It was an example sentence crafted by the losing side to showcase non-standard characters clustered in a single line. The reason this example endured for so long is clear: reading just one line makes the problem immediately obvious.

The counterarguments from those who defended the standard are also worth hearing. A surviving rebuttal stated, "Since the establishment of the completion-type standard, we have applied it to actual work and encountered no inconvenience whatsoever." Even an author on the opposing side conceded in the very same article that for administration, general office work, and the natural sciences, that statement was true. There was also an argument regarding telecommunications: that the protocols of international telecommunication lines made communication difficult with the combination-type code.

We can weigh that rebuttal once again. The telecommunications argument was not without merit. As we saw earlier, across 3,724 modern syllables, the combination code contained byte values that overlapped with the ASCII range, and its layout clashed with international protocols. Yet in July 1989, there was also a live demonstration where a portable computer running the combination code successfully communicated over international telephone lines from Sydney to Seoul. And the rebuttal that "you can simply make up for it with user-defined areas" ultimately breaks data interchange, as we saw in the previous chapter. Neither side's arguments collapse entirely, and neither side's arguments stand completely unscathed.

Attempts continued to fill in the missing letters. In 1991, a supplementary code set was established, adding 1,930 syllables to the main body's 2,350 syllables to make 4,280. Because that was still not enough, another 2,376 syllables were added to create a set of 6,656 syllables, which was reported to international organizations. Even with this set, however, 4,516 modern syllables were still missing. Neither internationally nor at home did this expansion plan ever gain traction.

Meanwhile, the market filled the void in a different way. An extended layout emerged that expanded the pre-composed (Wansung) list all the way to 11,172 syllables, and it quickly became widespread on personal computers. It worked by simply cramming the 8,822 syllables outside the standard main body into whatever unused slots were left.

This came at a cost. First, because of those newly crammed slots, the code values fell out of sync with alphabetical order. In this layout, "똠" came before "가". The supposed benefit—that alphabetical sorting came for free—vanished right here. Second, in this expanded layout, 3,606 syllables had one of their two bytes land in the range used by English letters. It was the exact same problem that plagued Johab's 3,724 syllables. When using only the standard 2,350 syllables, there had not been a single such character. Expanding the space brought back the very problems they had avoided. It laid bare the fact that the two-byte container was simply too small from the start.

There was movement on the policy front as well. In October 1992, Johab was finally accepted into the national standard. It was a decision reached after around ten committee meetings—including the Character Code Research Sub-committee—and a final public hearing on the 6th of that month. But there was a catch. The place Johab was given was an annex, and the definition of the 94×94 set in the main body used for information exchange remained untouched. An international standard document published in December 1993, which defined how Hangul was encoded in internet emails, nailed this point down: the definition was identical across the 1987, 1989, and 1992 editions.

In other words, the pathway leading to the outside world was still occupied by the main body. Johab had its name listed inside the national standard, but along the roads where documents and emails actually traveled, it never found a seat.

There is a third standard in this story: the universal character encoding system that aimed to fit all the world's writing into a single set. Its first version came out in October 1991.

Looking back, the premise at the starting line is striking. The architecture was designed to allocate a fixed 16 bits per character, accommodating just over 60,000 characters. That was written into the founding organization's original mission statement, a sentence that has since been revised to read that it has "surpassed 16 bits." In the recollections of someone who took part in early reviews, there is a line noting that "when we counted the upper limit of commonly used modern characters, they fit comfortably within 16 bits." Both standards had shared the premise that only frequently used characters needed to be included. And both ended up walking back that assumption later on.

If you look at the character list of Version 1.1 released in 1993, there is an intriguing detail. In the Hangul syllables block, there are 6,656 characters. It is the exact number from the previous chapter. The Extended Wanseong set that South Korea had reported to the international standards body had been brought directly into the universal character system. The initial ordering also followed the text of the domestic standard. In other words, the Hangul that the universal character system first accepted was the set from the side that had won domestically.

Then in July 1996, Version 2.0 wiped out that block entirely. It erased the 6,656 characters and laid out all 11,172 modern syllables in a brand-new space. A single line in that version's character database change log reads: "Deleted the old Hangul syllables and provided a mapping table to the new characters." Thousands of characters that had already been assigned code points were moved to completely different numbers. In character encoding, accidents rarely get bigger than this.

Yet what came next brings balance to this chapter. The universal character system now has a document called the Stability Policy. Its very first line reads: "Once a character is encoded, it will not be moved or removed." And the release where this policy takes effect is starting with Version 2.0. The policy never to change character names begins with that exact same version. The very release that made this massive move was the one where the promise never to move again began. Flaws did not belong solely to the losing standard; the winning standard, too, reached where it stands today only after turning its own flaw into a rule.

What fills this new space is the final twist of our story. If you open the text of today's global character encoding standard, you will find the name given to the block containing all 11,172 modern Hangul syllables: "the Johab set." The standard even notes where this block came from: drawn as a complete Johab set from the 1992 South Korean national standard.

And it is not just in name. The standard's text includes a list of constants: the starting code point of the syllable range, the starting offsets for initial consonants, medials, and final consonants, along with 19 initials, 21 medials, and 28 finals. By subtracting the base value from a syllable number, dividing, and taking the remainder, you derive the initial, medial, and final letters; conversely, calculating from the jamo yields the syllable number. The standard codifies this arithmetic as a normative rule and does not write out the breakdown of each syllable in a lookup table. It means that the 11,172 characters are not an arbitrary list in a table, but values directly derived from jamo.

The jamo used for combination also have their own dedicated space. There were 240 characters in version 1.1, and after archaic Hangul letters were added in version 5.2 in 2009, the count stands at 357 today. The standard explicitly notes that while every precomposed syllable code can be decomposed into a sequence of jamo, the reverse is not true. That is because there are thousands of archaic Hangul syllables that can only be written using combining jamo—syllables containing letters that modern orthography phased out of contemporary spelling.

Yet the principle of combination was not the only thing preserved. Precomposed syllables still hold their numbered spots, arranged in strict alphabetical order. The standard notes that "a binary comparison yields the sorting order used in the Republic of Korea." The advantage that had been broken in Unified Hangul Code (Extended Wansung) was revived here. In effect, both approaches came to coexist side by side within a single encoding system. In South Korea, a regulation establishing this encoding system as a national standard was officially promulgated in December 1995. Because it followed the International Organization for Standardization specification, it already included both the 240 jamo and the 11,172 precomposed syllables.

There is one more striking detail. The standard even documents how to construct the representative glyph of a syllable from its constituent jamo. If the medial vowel consists of vertical strokes, the initial consonant is placed to its left; if horizontal, on top; and if a hybrid of both, above the horizontal stroke and to the left of the vertical one. If there is a final consonant, it sits centered beneath the cluster. This does the exact same work as the Hapjahahe (Explanation of Combining Letters) in the Hunminjeongeum Haerye, which directed writing initial, medial, and final letters from left to right and top to bottom in syllabic blocks. Because Hangul was designed from the beginning for assembly, it meshed seamlessly with an encoding system governed by rules of assembly.

Today, a web statistics page tracking character encoding usage displays UTF-8 at 99.0%. That was the figure recorded on September 6, 2026, while Korea's legacy encodings appear down in the sub-0.1% list. Because the survey updates daily, the numbers will look slightly different the next time you open it. Yet the direction is unmistakable. The encodings that once contested this arena have now almost entirely vacated the stage.

Even so, there are still places worth watching. The first is legacy data. Documents and databases stored under older encodings still linger all over the place. Whether that file carries metadata noting which standard should be used to read it decides the lifespan of those records. The broken characters mentioned at the very beginning of our story resurface right here.

The second is the price of the rule that an assigned position never moves. Thanks to this principle, today's code point will continue to point to the exact same letter for generations to come. In exchange, misplaced spots must remain exactly as they are, because what was established in an earlier version cannot be undone in a later one. How we should view this trade-off remains an ongoing debate.

The third is how Old Hangul is actually handled. Archaic characters can only be written by lining up individual combining jamo. Yet inside the rules computers use to normalize text, there is a step that merges jamo back into precomposed syllable blocks. Combinations that form modern syllables merge together, while those that do not remain as loose jamo. As the digitization of classical literature grows, this boundary line will demand frequent hands-on care.

The list decided in 1987 has vanished, and the Johab encoding placed in the 1992 annex has vanished as well. What remains is not an answer to which side was right, but the fact that there was simply no answer within two bytes. Only after the vessel grew larger could both approaches sit side by side. If you erase a character on your screen right now, the entire square block disappears at once. Yet deep inside that block, the calculation that combines individual jamo is still running.