I have a strange phenomenon:

I'm translating from html to asciidoc, and would like to have auto_identifiers with headers in the output.

The html reader seems to enable auto_identifiers by default, as I can see those when I choose native as the output format (and don't assume 'native' will add them by itself).

In the asciidoc output though, the identifiers are only retained when I have an image as part of the header. I tried several alternative headers, and the identifiers only show in the output when the header contains an image.

Luckily, I can force consistent auto identifiers for all headings when using `-t asciidoc-auto_identifiers` on the command line (but mind the `-auto_...` instead of `+auto_...` to enable it!). But without that, why would the asciidoc writer automatically output the auto identifiers only for headers with an image? Is there some kind of complexity metrics involved?

Thanks, T.