Chinese telegraph code, but use UTF8 as a fallback.
CTC is the shorthand for "Chinese Telegraph Code".
It compiles most of the common characters and assigns them a number.
For example, character 汉 is numbered 3352.
This is a more compact way to index Chinese:
It only uses two bytes (actually it's only 0000 to 9999),
while UTF-8 uses three bytes on average.
While UTF-8 is generally the best option to handle all kinds of characters, when talking about encoding along with a limitation on length, CTC becomes useful.
This program is written for encoding text on postcards, while we don't want to deal with encrypted data represented in base58, but still want some level of privacy. So this program encodes the Chinese into CTC, and if there is a missing character, it uses either raw UTF-8 or compressed UTF-8 to encode.
Handwriting and typing numbers are much easier than base64/58.
The design is: when writing postcards, I encode the text either from cli options,
or files, or from stdin, then it spills out a bunch of numbers, and I just need
to write them on the postcard.
When someone got the postcard, he can type in the numbers and the program will print out the decoded message.
The program is built using jlink with zulu 21 jdk, so technically you don't need
a Java runtime installed on your pc.
Download the correct build for your usage and invoke ./ctc, everything should be fine.
There is no package published to use this codec as a dependency. I didn't write to dependency, I didn't test, so I don't want to people use it and accidentally cause some disaster. You can try with the jitpack, but I won't ensure the result.
If this demand is popular, create an issue and I will work on it to publish as a dependency
You can encode text from the cli option, file, and stdin.
./ctc encode -t "天匠染青红，花腰呈袅娜。"
./ctc encode -f message.txt
echo "天匠染青红，花腰呈袅娜。" | ./ctc encode -c UTF-8The -t <text> option will omit the -f <file> if they are both specified.
When using the stdin, you need to manually tell the program you're done by inputting
EOF on a new line.
You can change the EOF to something else using --eof=EOT, now you can use EOT
instead of EOF.
You can also change how many codes are printed in one line.
But do be aware this is limited by your terminal width, if it's longer than the width, it will wrap.
By default, it's fit your terminal.
You can change it by option -w 6, now you have six codes per line.
When using stdin input, there is some issue with the charset on Windows.
By default, the program uses UTF-8 to read and write everything.
But on windows, it will use GBK if the default locale is zh-CN;
or use BIG5 if the default locale is zh-TW.
If you don't want this feature, you can use -c UTF-8 to force the program to
use UTC-8 charset.
Do verify the program output before you write numbers to your postcard. Sometimes the charset issue can mess you up if you don't check the decoded result.
It's basically the same, but much simpler. When decoding, the program will print everything using UTF-8, which most terminals are ok, even on windows.
For decoding, there are only three options: -t, -f and --eof.
You can type any character, and the program will ignore them and only take the digits. So don't worry if you accidentally typed a letter or something. Keep typing numbers and you're good.
./ctc decode -t "1131 0561 2676 7230 4767 9976 5363 5212 0701 5934 1226 9975 84"Starting from v1.0.0, I removed the old transposition cipher because it's not safe. Instead, I added ChaCha20 stream cipher for encryption.
The ChaCha20 stream cipher will generate a stream of fake codes ranged from 0000 to 9999 (key). Then for each encoded CTC code, we calculate the encrypted code:
enc = (code + key) % 10000
When decrypting, the same stream of fake codes will be generated using the same key and nonce:
code = (enc - key + 10000) % 10000
The nonce will be generated at the encrypting time,
and is required via cli option --nonce (-n for short) during the decrypting time.
The key (-k or --key) will be decoded differently depending on the encryption mode.
If you use symmetric encryption, the key will be decoded as UTF-8 bytes and apply SH3-256. The result will be the ChaCha20 stream key.
If you use asymmetric encryption, the key will be decoded as hex string to be private key,
then the --dh hex option will be used to as your partner's public key (also as hex string).
Then use X25519 ECDH to generate the shared secret, and then apply SHA3-256.
The result will be the ChaCha20 stream key.
For asymmetric encryption, to generate the private key, use openssl rand -hex 32.
To get the public key, use ./ctc pubkey <your private key in hex> to calculate the public key.
When encryption is enabled, the program will also generate an MAC to ensure the integrity of the message.
This will be generated on encrypting time, and will be required on decrypting time via --mac hex (-m for short).
There is some internal processing beyond simply lookup the table and do the substitution. The CTC is scraped from https://en.wiktionary.org/wiki/Appendix:Chinese_telegraph_code/Mainland_1983.
Before starting encoding characters, we need to replace some text for shorter codes:
- new lines (
\n) are replaced by〷(code 9999) - text like
10日will be replaced by the char㏩(code 9910)
This will save some code and will be transparent for user. When decoding, reversed processing will be applied to make sure the decoded text is looking normal.
When there is an unknown character, it will save to the buffer and encode them in batches.
For example, when encoding 大家好，my name is XXX。, the unknown buffer will be my name is XXX,
then it will be encoded to utf-8 bytes.
The encoded bytes will be compressed by LZMA2 with preset max and dict size of 256KB,
and the result will be kept if the compress result is shorter than the raw form.
For raw UTF-8 bytes, it will be enclosed with 9992 and 9993, the content
will be representing in 0000 to 9991, like base9992 but all in numbers.
For compressed UTF-8 bytes, it will be enclosed with 9994 and 9995.
The content representation is the same.
To ensure the integrity of the encoded data and also to ensure the typing is correct, I added a checksum at the end of each encoded line.
For example, the text 天匠染青红，花腰呈袅娜。 corresponding to the following code:
1131 0561 2676 7230
4767 9976 5363 5212
0701 5934 1226 9975
When printing out each line, the ISO 7064 mod 97–10 checksum will be appended to the end.
So the actual output will be (use -w 4 to specify 4 codes per line):
1131 0561 2676 7230 17
4767 9976 5363 5212 92
0701 5934 1226 9975 62
When decoding, the checksum is always REQUIRED. If you don't know/have the checksum,
or the checksum is blurred and is indistinguishable, just use 00 (must be two digits).
Also, the line break is also required since the decoding process will check each line separately.
So you can't mix two lines into one. You have to type them as-is (blank chars are ok, they will be ignored).
Welcome, but I can't guarantee if I accept the pr or not. Open issues as long as you have questions, but I can't guarantee to answer them.
TL;DR: I give no guarantee on the program, unless you pay me.