Why do we have so many ways to create pdfs, and why do most libraries/frameworks only provide a partial solution? At first it seems not to difficult to build something for what I need, so it makes sense to build it myself. But the pdf spec is large, and I don't use most of it. Building it myself makes more sense than trying to understand a current implementation. That is more complex than needed to solve the current problem, and building my own is a good way to understand the spec. Completing an implementation is hard to justify on project costs (as is mostly done for open source). More people seem to have followed the same road, so we end up with lots of partial solutions. Next time I need to create pdfs, I have different needs. Something with fonts, that seems like a lot of work. So I connect to a c library that already implements it. Only that has no way to create a document in-memory, just on disk… Artefact at the moment has very simple text layout and table algorithms. They are suitable for simple documents. High quality layout is complex as it needs to handle a.o. line breaking, justification, ligatures, kerning, hyphenation, text direction, text container shape, paragraph and page layout. PDF4Smalltalk is large because it is able to read and write pdfs, so it needs to have at least basic support for all pdf elements, and it contains lots of constants and mappings (glyph list, font metrics). Cairo (and Pango) by itself does not support reading pdfs. Stephan