update user guide part of the book

book/src/SUMMARY.md

@@ -15,6 +15,7 @@
   - [#[global_logger]](./global-logger.md)
   - [panic! and assert!](./panic.md)
   - [Printers](./printers.md)
+  - [Encoding](./encoding.md)
 - [Design & impl details](./design.md)
   - [Interning](./interning.md)
     - [Dealing with duplicates](./duplicates.md)

book/src/encoding.md

@@ -0,0 +1,32 @@
+# Encoding
+
+> 💡 Most users won't need to change the encoding so this section is mainly informative.
+
+`defmt` data can be encoded using one of these 2 formats:
+
+- `rzcobs` - [Reverse Zero-compressing COBS encoding][rzcobs] (rzCOBS). This is the default encoding.
+- `raw` - raw data, that is no encoding.
+
+[rzcobs]: https://github.com/Dirbaio/rzcobs
+
+In comparison to not using any encoding, `rzcobs` compresses the data (uses less transport bandwidth),
+and adds some degree of error detection thanks to its use of frames.
+
+The encoding is selected via a Cargo feature on the `defmt` crate.
+These Cargo features are named `encoding-{encoder_name}`, e.g. `encoding-rzcobs` and `encoding-raw`.
+
+``` toml
+[package]
+name = "my-application"
+
+[dependencies.defmt]
+version = "0.3.0"
+features = ["encoding-rzcobs"] # <- encoding
+```
+
+Note that libraries (dependencies) MUST not select the encoder so that applications (top-level crates) can.
+If no `enocding-*` feature is enabled then the default encoding is used.
+
+The encoding is included in the output binary artifact as metadata so [printers](printers.html) will detect it and use the appropriate decoder automatically.
+When the `rzcobs` encoding is used the printers will skip malformed frames (decoding errors) and continue decoding the rest of the `defmt` data.
+In contrast, printers handling the `raw` encoding will exit on any decoding error.

book/src/introduction.md

@@ -8,7 +8,7 @@
 
 - `println!`-like formatting
 - Multiple logging levels: `error`, `info`, `warn`, `debug`, `trace`
-- Crate-level logging level filters
+- Compile-time `RUST_LOG`-like filtering of logs: include/omit logging levels with module-level granularity
 - Timestamped logs
 
 ## Current limitations

book/src/istr.md

@@ -11,7 +11,7 @@ A `Str` value is created using the [`intern!`] macro; the argument to this macro
 # extern crate defmt;
 let s = "The quick brown fox jumps over the lazy dog";
 defmt::info!("{=str}", s);
-//                      ^ bandwidth-use = 43 bytes
+//                     ^ bandwidth-use = 43 bytes
 
 # use defmt::Str;
 let interned: Str = defmt::intern!("The quick brown fox jumps over the lazy dog");
@@ -21,12 +21,12 @@ defmt::info!("{=istr}", interned);
 
 > ⚠️ This was a contrived example to show the difference in bandwidth use.
 >
-> In practice you should following, which also interns the log string and uses as little bandwidth as the `{=istr}` version:
-> 
+> In practice you should use the following, which also interns the log string and uses as little bandwidth as the `{=istr}` version:
+>
 > ``` rust
 > # extern crate defmt;
 > defmt::info!("The quick brown fox jumps over the lazy dog");
 > ```
 
 [`defmt::Str`]: https://docs.rs/defmt/*/defmt/struct.Str.html
-[`intern!`]: https://docs.rs/defmt/*/defmt/macro.intern.html
+[`intern!`]: https://docs.rs/defmt/*/defmt/macro.intern.html