1 files changed, 43 insertions, 0 deletions

diff --git a/19.md b/19.md
new file mode 100644
index 0000000..9235196
--- /dev/null
+++ b/19.md
@@ -0,0 +1,43 @@
+NIP-19
+======
+
+bech32-encoded entities
+-----------------------
+
+`draft` `optional` `author:jb55` `author:fiatjaf` `author:Semisol`
+
+This NIP specifies all bech32-encoded entities.
+
+## Bare keys and ids
+
+To prevent confusion and mixing between private keys, public keys and event ids, which are all 32 byte strings. bech32-(not-m) encoding with different prefixes can be used for each of these entities.
+
+These are the possible bech32 prefixes:
+
+  - `npub`: public keys
+  - `nsec`: private keys
+  - `note`: note ids
+
+Example: the hex public key `3bf0c63fcb93463407af97a5e5ee64fa883d107ef9e558472c4eb9aaaefa459d` translates to `npub180cvv07tjdrrgpa0j7j7tmnyl2yr6yr7l8j4s3evf6u64th6gkwsyjh6w6`.
+
+The bech32 encodings of keys and ids are not meant to be used inside the standard NIP-01 event formats or inside the filters, they're meant for human-friendlier display and input only. Clients should still accept keys in both hex and npubformat for now, and convert internally.
+
+## Shareable identifiers with extra metadata
+
+When sharing a profile or an event, an app may decide to include relay information and other metadata such that other apps can locate and display these entities more easily.
+
+For these events, the contents are a binary-encoded list of `TLV` (type-length-value), with `T` and `L` being 1 byte each (`uint8`, i.e. a number in the range of 0-255), and `V` being a sequence of bytes of the size indicated by `L`.
+
+These are the possible bech32 prefixes with `TLV`:
+
+  - `nprofile`: a nostr profile
+  - `nevent`: a nostr event
+
+These possible standardized `TLV` types are indicated here:
+
+- `0`: `special`
+  - depends on the bech32 prefix:
+    - for `nprofile` it will be the 32 bytes of the profile public key
+    - for `nevent` it will be the 32 bytes of the event id
+- `1`: `relay`
+  - A relay in which the entity (profile or event) is more likely to be found, encoded as UTF-8. This may be included multiple times.