Possible replacement for EPUBCHECK

[PeterT]

This might be of interest to some.

A new tool epubverif is under active development as an alternative to the Java based epubcheck. See https://github.com/veripublica/epubveri

I've no association with the project but thought it might be of interest. I did do a download and build under Windows 11 and it seemed fine.

Why does this exist? Isn't epubcheck already the standard?


Yes — epubcheck is the official, W3C-maintained EPUB validator, and it's the industry standard. Every major publishing pipeline uses it somewhere. epubveri is not trying to discredit it or replace its authority — it's trying to solve one specific, practical problem with it:

epubcheck is written in Java, and needs a JVM (Java Virtual Machine) to run. That's fine if you're running it as a standalone command-line tool on a server. It's a real obstacle if you want to:

run validation inside a web browser (there's no JVM in a browser — but there is WebAssembly, which Rust compiles to natively),
embed a validator inside a native desktop/mobile app without bundling an entire Java runtime,
run validation as part of a fast, lightweight command-line tool or CI pipeline without the JVM startup cost, or
just avoid the operational overhead of "make sure a compatible JVM is installed" as a deployment requirement.
epubveri is written in pure Rust, with zero C dependencies and zero JVM dependency. Rust code compiles down to a small, fast native binary — and, crucially, can also compile to WebAssembly (WASM), meaning the exact same validation logic can eventually run directly in a web browser tab, with no server round-trip and no Java anywhere.

It reuses epubcheck's own short error-code scheme (OPF-002, RSC-005, PKG-007, and so on) wherever the checks overlap, specifically so that anyone already familiar with epubcheck's output — or any tooling built to parse it — recognizes epubveri's output immediately. This is a deliberate compatibility choice, not a coincidence.

[DNSB]

I tried it out. First issue is that on multiple ePubs where the epubcheck v5.3.0 did not indicate any errors, epubveri came up with the following errors:

C:\Users\<user>\Desktop>D:\builder\epubveri\target\release/epubveri monospace_test.epub
ERROR RSC-005: value of attribute "property" is invalid (must not be empty) [OEBPS/content.opf]
INFO OPF-088: epub:type value 'frontmatter' is not in the default vocabulary [OEBPS/Text/nav.xhtml]
ERROR OPF-096: non-linear content 'OEBPS/Text/nav.xhtml' is not reachable from the reading order [OEBPS/content.opf]
— 2 error(s), 0 warning(s): INVALID

Near as I can tell, the only property in the .opf file is not empty(*) and nav.xhtml is properly flagged as non-linear and so should not be accessible from the reading order. And the info about frontmatter? Near as I can tell, it is one of the default items according to the latest ePub3 spec under section 2: Document Partitions.

<meta property="dcterms:modified">2026-03-02T09:56:05Z</meta>

I tried a couple of other ePubs and got a mix of valid and invalid error messages. Some of the messages were ones that earlier versions of epubcheck would generate but all the ones I am aware of were fixed in epubcheck v5.3.0 back in Sept., 2025.

I suspect I should have dumped my old Rust install. It took more effort to update than a simple install would have taken.

[JSWolf]

I see no need for epubverif.

[rantanplan]

Originally Posted by JSWolf

I see no need for epubverif.

Not having to install Java is always a plus so even though there's technically no need for it, I would prefer it to a Java jar.

EDIT
And before I forget, epubverif doesn't seem to be vibe coded, so that's a plus for you right there.

[KevinH]

And based on the error output snippet shown by DNSB, it is not providing any line and column number for any of the errors making it a non-starter for use in Sigil as a validator.

Without a rough position for each error, it is not useful.

And Rust outside of a web/kernel/security focused environment is also a bit of a pain in the ass. One more toolchain, one more runtime library, etc. And Rust source is much harder to read/follow than java for anyone not already familiar with it.

So it needs to at least add error position info to be at all useful, and will need to fix itself to exactly match epubcheck (fix their bugs) before we can consider using it inside Sigil.

And frontmatter is clearly listed in the epub structural semantics spec and is legal:
See: https://www.w3.org/TR/epub-ssv-11/

[JSWolf]

And epubcheck works very well with sigil and calibre. Why change what already works so well?

[KevinH]

Created an issue on their github site pointing here. So if you find any more issues or differences from epubcheck that you feel it would be useful for the epubverif devs to know about, please post them here, or better yet, create proper issues in their repo if you have a github account.

[KevinH]

Originally Posted by JSWolf

And epubcheck works very well with sigil and calibre. Why change what already works so well?

Because options are good. Sigil will never abandon epubcheck but just like the old flightcrew before it was retired, having multiple validators is a good thing if they catch different but real issues.

[rantanplan]

Originally Posted by JSWolf

And epubcheck works very well with sigil and calibre. Why change what already works so well?

Why do you update your Calibre if the previous version works so well? Why did you ever move on from 7.1.3?

[KevinH]

FWIW, the epubverif devs are quite responsive to our bug reports, and have already made improvements:

https://github.com/veripublica/epubveri/issues/1

So if anyone is willing to work with them to find more issues (after updating to to the latest version), they would be happy to hear from you!

Please consider helping them out.

[JSWolf]

If the devs want my help, they'll have to release a compiled version. I'm not installing Rust to compile every new version.

[PeterT]

Originally Posted by JSWolf

If the devs want my help, they'll have to release a compiled version. I'm not installing Rust to compile every new version.

Seeing the devs are not here I guess you'll need to contact them via github

[PeterT]

Originally Posted by JSWolf

If the devs want my help, they'll have to release a compiled version. I'm not installing Rust to compile every new version.

Looks as if there are binary builds available at https://github.com/veripublica/epubveri/releases

[DNSB]

Originally Posted by PeterT

Looks as if there are binary builds available at https://github.com/veripublica/epubveri/releases

Looking at the binary build from https://github.com/veripublica/epubveri/releases and the one I compiled, I am no longer getting the frontmatter info message but the two error messages are still showing up.

[KevinH]

Hopefully with line and column information now.