The http and https references were anyway "accidents" and they should have been anyway removed. It's not messing with the book but rather a natural evolution. I am the author and I decide what it contains and how it displays it.
There are no errors, only the same things repeated on and on for each of the files in manifest (parsing bla bla bla). It was a long file, for I have more than 200 text files and even more images. And it was discarded anyway.
The problem newer versions have and the main reason I keep some older ones is that newer versions badly screwed up the
Calibre official way of printing an epub, via PDF. Why I need printing? For I can read better on real paper and see the spelling errors or any other errors (for instance, I can compare two pages for duplicated ideas or statements and stuff, and I can better see whether a certain glyph is lacking from a font, which I can't always see in the preview window).
I am sorry that I thought it would be an easy job, for I never thought a DNS error should occur in an epub->pdf conversion
, particularly when the author said that accessing the external resources was a
corrected error of some earlier versions, and therefore if occurred it should have occurred in a specific module. From what I learned yesterday, it may be one of the errors passed to calibre by chrome pdf-engine.
Anyway, the problem was solved by using the conversion of a good old 2.5 version.