diff --git a/pdf/PDF.pdf b/pdf/PDF.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..03f08cd9504070a5b38da33745aae3b3977b055b
Binary files /dev/null and b/pdf/PDF.pdf differ
diff --git a/pdf/pdf.html b/pdf/pdf.html
index 068a68e2a12adb44841a8cbe6269fa641b28a33f..9f34d859223fb87e2859294c24497395da35e71e 100644
--- a/pdf/pdf.html
+++ b/pdf/pdf.html
@@ -232,14 +232,7 @@ endobj
 
 ---
 
-# Content
-
-Text (`w` = line width, `Tf` = font, `TL` = leading, `Tj` = text)
-```html
-stream
-BT 1 0 0 1 22 20 cm 1 w /F 12 Tf 14.4 TL (Hello world)Tj ET
-endstream
-```
+# Drawing
 
 Drawing (`rg` = background color, `re` = rectangle, `b` = painting mode)
 ```html
@@ -248,7 +241,7 @@ stream
 endstream
 ```
 
-<img src="images/rectangle.png" width="20%">
+<img src="images/rectangle.png" width="30%">
 
 
 ---
@@ -305,7 +298,7 @@ class: center, middle
 
 # Text with default encoding / fonts
 
-Can use embedd default fonts (`Helvetica`) with default encoding `WinAnsiEncoding`.
+Choose predefined font (`Helvetica`) & predefined encoding (`WinAnsiEncoding`).
 
 ```html
 4 0 obj
@@ -316,7 +309,7 @@ endobj
 endobj
 ```
 
-Then print text like:
+Then print text like (`w` = line width, `Tf` = font, `TL` = leading, `Tj` = text)
 ```html
 6 0 obj
 <</Length 59>>
@@ -434,18 +427,25 @@ What about suficient?
 | hmtx | width of glyphs |
 | glyf | list of glyph data blocks |
 
-In short: Read out tables, patch them together differently, hope offsets work.
-
 ---
 
-# Subsetting
-
-Fonts usually have 100's of glyphs.  
-Do we need more than 60?
+# Overview `ttf` processing
 
+Binary read out:
+- Read out table directory to get offset / type of tables
+- Read out each table (possibly multiple formats; `cmap` has 4!)
 
+Interpret:
+- Split `glyph` table as mandated by `loca`
+- Get character widths as defined by `hhea` and `maxp`
+- Use `post` and `cmap` to get unicode per `glyph`
 
+Write back:
+- Create subset with `glyphs` we actually need
+- Correct "summary" tables (like #characters in font)
+- Recreate binary file
 
+In short: Read out tables, patch them together differently, pray everything works.
 
 ---