diff --git a/content/cn-date.md b/content/cn-date.md
new file mode 100644
index 0000000..ddbc5cb
--- /dev/null
+++ b/content/cn-date.md
@@ -0,0 +1,119 @@
+---
+title: chinese date format
+---
+
+## Format
+
+* Traditional Chinese: `二零二零年四月二十八日`
+* Simplified Chinese: `2020年04月28日`
+
+## Characters
+
+| Character | Meaning |
+| --------: | ------: |
+|        零 |       0 |
+|        〇 |       0 |
+|        一 |       1 |
+|        二 |       2 |
+|        三 |       3 |
+|        四 |       4 |
+|        五 |       5 |
+|        六 |       6 |
+|        七 |       7 |
+|        八 |       8 |
+|        九 |       9 |
+|        十 |      10 |
+|      二十 |      20 |
+|      三十 |      30 |
+|        年 |    Year |
+|        月 |   Month |
+|        日 |     Day |
+
+## jq implementation
+
+I implemented a chinese date parser using [jq][jq] for [itsb][itsb], as seen
+[here][helpers].
+
+The parsing methods are in two parts: `parse_chinese_number`, a parser that is
+only guaranteed to work from 0 to 99, and `parse_chinese_date`, which splits
+the date components and sends them to `parse_chinese_number`.
+
+This parser only works with jq≥1.6, as jq 1.5 and earlier had some
+[Unicode issues][unicode] that caused most string manipulations in this parser
+to break.
+
+### Number parsing
+
+In Chinese dates, years are always expressed using all of their digits, aka
+"two zero two zero" for 2020, and not "two thousand twenty".  This makes the
+parsing much simpler as I do not need to even know how thousands or hundreds
+are expressed in Chinese; I still however need tens to handle months and days.
+
+I first started by writing a parser that only handles single digits; I made an
+object mapping Chinese characters to their string digits, and just translated
+each character then re-concatenated.
+
+```
+($input // "")     # ["二", "零", "二", "零"]
+| map($charmap[.]) # ["2", "0", "2", "0"]
+| join("")         # "2020"
+```
+
+I then added 10 as an empty string, because we can just ignore it when going
+number by number.  This only works in some cases:
+
+| Number | Parsed as | Expected |     Actual |
+| ------:| ---------:| --------:| ----------:|
+| 二十八 |      二八 |       28 |         28 |
+|   十八 |        八 |       18 |          8 |
+|   二十 |        二 |       20 |          2 |
+|     十 |      `""` |       10 | Type error |
+
+I chose to ignore any case where more than one 十 would be found as my goal
+was only to parse in the 1-31 range, and 十十十一 is longer than 三一 or
+三十一 so I can expect them to not be used.
+
+The remaining edge cases only occur when `十` is at the start or the end of
+the string, so I handled them in three ways:
+
+* If the string is exactly `"十"`, return 10 before any other parsing;
+* If the string ends with 十, multiply the output by 10;
+* If the string begins with 十, add 10 to the output.
+
+And to avoid adding complexity to my feed parsing scripts, I changed the
+`map` to `map($charmap[.] // .)`, which just ignores unknown characters.
+Combined with some checks made by the date parsing function, this makes it
+possible to parse both traditional and simplified formats without making
+many changes.
+
+### Date parsing
+
+The date components are split using a regex, to avoid sending too much garbage
+to `parse_chinese_number` in the event of a badly formatted date. Once the
+numbers are parsed, we get an object in this format:
+
+```json
+{"year": 1234, "month": 12, "day": 3}
+```
+
+In some cases, some investigation agencies will be using years from the Chinese
+calendar, where year 0 is year 1911 of the Gregorian calendar.  I therefore
+added a check that adds 1911 years when the year is below 1900; this causes the
+parser to work properly only for years between 1900 and 2811 inclusive for
+dates using this calendar.
+
+I then use a rather simple method to get a Unix timestamp:
+`"\(.year)-\(.month)-\(.day)T00:00:00Z" | fromdateiso8601`.  I could have used
+a more normal method such as a `strptime("%Y-%m-%d") | mktime`, or build the
+same array that `strptime` returns, such as
+`[1234, 11, 2, 0, 0, 0, 0] | mktime`, but that requires some particular
+handling as months and days are zero-based in this format.
+
+## Acknowledgements
+
+Thanks to ~m455 for the Chinese dates crash course on IRC!
+
+[helpers]: https://tildegit.org/lucidiot/itsb/src/commit/70ef6b9e978aa58170e80bbb7121f3a6089f8ac3/jq/helpers.jq#L45
+[itsb]: https://tilde.town/~lucidiot/itsb/
+[jq]: https://stedolan.github.io/jq/
+[unicode]: https://github.com/stedolan/jq/issues/1166