epubcheck, any software can pass its validiation check?

[droople]

Hi

I just used epubcheck to validate some epub files made by Calibra, eCub, Sigil, quite interesting, none of them can pass the validation check, although all the books can be read on the machine.

Does this mean that actually no one care about the epub specification?

Cheers

PS: I found the epub file made by Atlantis can pass the validation check.

[charleski]

A lot of files fail because they're using xhtml that doesn't precisely comply with the relevant specification. Most of these errors are trivial at the moment, but may become non-trivial in the future as the xhtml spec evolves.

The purpose of the spec is to make sure that you'll be able to read your epubs on future devices as well as the ones you have now, so yes, it's a good idea to make sure they're compliant.

[Valloric]

Originally Posted by droople

I just used epubcheck to validate some epub files made by Calibra, eCub, Sigil, quite interesting, none of them can pass the validation check.

If an epub you saved with Sigil doesn't pass epubcheck, then it's usually because the XHTML imported into Sigil had problems. If you see an instance where Sigil saves an epub that doesn't pass epubcheck where the original file did, then that's bug report (of high priority).

I try to make Sigil epubcheck-clean; that is, I try my best to makes sure Sigil doesn't introduce non-compliance. But if the original file was non-compliant, there's a good chance Sigil won't be able to make it compliant (although it usually can).

[droople]

Hi Valloric
Thank you for the reply.

What I did with Sigil epub file is type in "test" in the text field, without import any external files.

Here is the validation message
ERROR: test.epub/OEBPS/content.opf(6): unfinished element

And please find the attached file.

Cheers

[kovidgoyal]

Just so you know: epubcheck is totally meaningless. A file that *passes* epubcheck may or may not work with any given epub renderer. A file that *does not pass* epubcheck may or may not work with any given epub renderer.

The best that could possibly be said about epubcheck (and I wouldn't be comfortable saying this without actual data) is that a file that passes epubcheck may be more likely than a file that does not pass to render correctly with most epub viewers. Even if this were true it would most likely be so because files that tend to pass epubcheck tend, on average, to have extremely simple markup, as they are typically the product of machine translation from some extremely simple format.

What I'm trying to say is that the things that epubcheck checks are those things that it is easiest to write software to check, not those things that are most likely to cause problems, or those things that are most likely to occur in the wild.

What epubcheck is good for, is those situations where you have absolutely no idea why your epub file is not rendering with a particular renderer. In that case, you can try running epubcheck on it and fix the errors it points out. Of course, that may or may not fix your actual problem. And even for this use case, epubcheck is extremely sub-optimal since its error messages are incredibly unhelpful.

That's my epubcheck jeremiad for this week.

[awp]

Originally Posted by droople

PS: I found the epub file made by Atlantis can pass the validation check.

Any EPUB file generated by Atlantis is supposed to pass the EPUB validation test. If you have a document that Atlantis Word Processor does not convert to a "valid EPUB", please email it to support@AtlantisWordProcessor.com.

Originally Posted by kovidgoyal

Just so you know: epubcheck is totally meaningless. A file that *passes* epubcheck may or may not work with any given epub renderer. A file that *does not pass* epubcheck may or may not work with any given epub renderer.

It is not completely meaningless. Some publishers do not accept EPUBs that do not pass the EPUB validation test.

[awp]

Originally Posted by droople

Here is the validation message
ERROR: test.epub/OEBPS/content.opf(6): unfinished element

And please find the attached file.

This is because some mandatory metadata items are missing.

[droople]

Originally Posted by kovidgoyal

Just so you know: epubcheck is totally meaningless. A file that *passes* epubcheck may or may not work with any given epub renderer. A file that *does not pass* epubcheck may or may not work with any given epub renderer.

The best that could possibly be said about epubcheck (and I wouldn't be comfortable saying this without actual data) is that a file that passes epubcheck may be more likely than a file that does not pass to render correctly with most epub viewers. Even if this were true it would most likely be so because files that tend to pass epubcheck tend, on average, to have extremely simple markup, as they are typically the product of machine translation from some extremely simple format.

What I'm trying to say is that the things that epubcheck checks are those things that it is easiest to write software to check, not those things that are most likely to cause problems, or those things that are most likely to occur in the wild.

What epubcheck is good for, is those situations where you have absolutely no idea why your epub file is not rendering with a particular renderer. In that case, you can try running epubcheck on it and fix the errors it points out. Of course, that may or may not fix your actual problem. And even for this use case, epubcheck is extremely sub-optimal since its error messages are incredibly unhelpful.

That's my epubcheck jeremiad for this week.

Oh, ok

[kovidgoyal]

Originally Posted by awp

It is not completely meaningless. Some publishers do not accept EPUBs that do not pass the EPUB validation test.

That just goes to show that those publishers don't have a clue.

[droople]

Originally Posted by kovidgoyal

That just goes to show that those publishers don't have a clue.

oh, ok.

[awp]

Originally Posted by kovidgoyal

That just goes to show that those publishers don't have a clue.

Maybe. But it is always safer to follow the specifications. If the OPF specification says that the "title", "identifier" and "language" metadata items are "required", why not include them in any EPUB file?

"Invalid" EPUBs might be accepted by the currently available EPUB readers. But there is no guarantee that they will be accepted by future readers.

[pdurrant]

It's a design decision not to add random values to the metadata for the required metadata fields.

If you use the metadata editing interface in Sigil, and enter values for the required items, Sigil ePubs pass epubcheck if you also have valid XHTML in the content.

Originally Posted by droople

What I did with Sigil epub file is type in "test" in the text field, without import any external files.

Here is the validation message
ERROR: test.epub/OEBPS/content.opf(6): unfinished element

[charleski]

Lol, I was expecting Kovid to take the chance to bang on about epubcheck again.

While it's not perfect (and desperately needs another release) I've found it very useful as a basic 'sanity check' to make sure that all the required elements are in the right places. Unfortunately many converters still output deprecated xhtml attributes, which is what causes most of the errors.

[kovidgoyal]

Originally Posted by awp

Maybe. But it is always safer to follow the specifications. If the OPF specification says that the "title", "identifier" and "language" metadata items are "required", why not include them in any EPUB file?

"Invalid" EPUBs might be accepted by the currently available EPUB readers. But there is no guarantee that they will be accepted by future readers.

Don't get me wrong, I'm not advocating that people ignore the standards on purpose. I just want people to realize that having your book passed by epubcheck is a guarantee of precisely nothing. I think having something like epubcheck does more harm than good, simply because everybody automatically assumes that if you have an EPUB that passes epubcheck's tests, it will work correctly.

epubcheck should have been named epub-schema-check to help make it clear that all it does is validate a few XML schemas.

[kovidgoyal]

Originally Posted by charleski

Lol, I was expecting Kovid to take the chance to bang on about epubcheck again.

Yeah, I just can't resist There's something about epubcheck's naivete that really gets me going.