Hello, VladD2, you wrote: VD> Tell you you do not see logical errors these lines? Well, how it is possible to prove any examples "" something? And generators of parcers, CSS, Markdown, HTML, LaTeX, Liquid? They as in your "theory" it is inscribed? Vlad, at me dab here is wider. I quite agree that Nitra approaches for handling exterior DSL, including set forth above. Simply I do an emphasis that internal it is much more convenient. VD> and the main thing! How worthless integration DSL th can justify application which is just created for good integration? Exterior DSL - worthless integration. In Nitra good integration only with N1. VD> Sounded XAML and QML from that also are ugly that their authors instead of creation of readable syntax decided to use XML, and in absence something comparable with Nitra washed down typification of these . In QML not XML, and a certain json-similarity with js, but you are absolutely right, typification there misses, as DSL completely exterior and as language was dynamics is taken certainly. But there convenient , instead of poverty from . VD> And at all thus programmers prefer to describe UI on XAML and QML, instead of the similar code on C# and a C ++. Because these languages are not suitable for creation DSL. But how helps here Nitra? You will not implement parser/tipizator for seamless integration. N>> the correct way - LINQ, that is seamless integration into target language. VD> aha. Only one way! Not only, but preferable. IRL all certainly is more difficult. VD> and on surprising coincidence, Nitra just also it is intended for creation of such languages. And it gives both means of seamless integration exterior, and internal DSL. Internal DSL there for N1 that not so . For exterior languages Nitra as I already told, approaches, but I do not know any seamless (with etc.) Integrations exterior DSL with target language. VD> the generator if it is based on such high-level abstractions (by the way, DSL) as Java byte code, LLVM or.Net MSIL is insignificant enough code amount in the compiler. The compiler master code is made by check of types and resolution of names. I do not have confidence that check of types which you implemented in Nitra strongly helps at implementation of the modern languages with the heaped up type systems, for example such as the Rock or Rast. The parcer on indents as I understand, too is not present. On all time etc. is necessary, but I simply judge about leaking a state and perspectives. VD> on what to create easier? In general purpose language, writing all stages of compilation manually, or on means which does not give only the last stage? In general I do not know, fairly. If language + a platform.+ knowledge of Nemerle => on Nitra probably faster. VD> to Tell "" and it "is not necessary" always is easier, than even simply to investigate in . I investigated into a problematics and anywhere you did not name. Simply area application and the approach to me does not see perspective.