[geek1011]

1. speed is not an issue for me. Yes, as an user sometimes I feel Calibre slow. Python *is* slow, but it's more simple to code and to maintain. A saw a lot of motivated programmers that refuses to code something that is not near to the metal, and they will soon abandon the project because it's too complicated to code, and no one will continue to maintain it for the same reason. Especially if it's written in that crap language that is C

It's not written in C. And note that FBInk, along with many other tools by many other people, is written in C. And, you'd find speed an issue if you often make bulk changes to books and factory reset Kobos often to test things.

2. a Calibre plugin is much more powerful, since it integrates not only with a mere converter, but with a full suite, thought to catalogue, add metadata and send books properly to the devices.

That is true, and I've never denied it. I've even pointed people to it instead of my own tools.

3. you can't know what will happen if you transform the HTML, even if you simply fix the syntax and remove useless tags. HTML renderers are various and unpredictable.For example, if you remove an apparently useless tag, you could break some css rule. The keyword is: option. You can add it as an option, maybe true by default, but not force it.

Calibre does more parsing/reencoding than kepubify. And, many of the transformations are made by specific regexps (it's better to miss a bit than to break the book).

4. don't take it personally, but I know dadivfor by a lot of time. He is skilled in epub and kepub format, Calibre, Kobo and many other e-ink related stuff. You... well, I read a few posts, but you seems to me very noob. The fact that you don't use Calibre plugin but a self-made solution, not configurable, and thinks that Calibre is a crap because is slow confirms my opinion. If you want a fast Calibre plugin, you can use Cython. Very simple and powerful. You don't need C++.

If you actually read my posts and code, you'd see I have nothing against Calibre itself (I actually think it is quite well-written, but a bit heavy), but it just doesn't fit my needs. I know a lot about EPUBs too (I've read the spec, participated in a few community group calls, and I have contributed to multiple EPUB-related projects).

I want to be clear, I have nothing against C++. It's a fantastic language, very powerful. But it's complicated, difficult to maintain, difficult to get help by someone else and it's prone to errors, memory leaks and security flaws if bad programmed. Python on the contrary is a pleasure to code, is readable, it's more simple to maintain and to avoid errors, and you can speed it up with Cython. Or with Python C API, if you want to be more near to the metal.

I don't use C++. I actually strongly dislike it, for some of the reasons you said yourself. But, note that the entire Kobo firmware is written in C++.