Hi Anthony!

> Attempting to infer semantic information (links) from a manpage written
> in an inherently non-semantic format like man(7) has poor prospects.

Technically correct. However for practical purposes it seems to work.


> A typical example is yacc(1);

Actually the yacc(1) man page is handled perfectly by my post-processing 
solution. But you are still correct for the general case.

I am currently processing <i>query</i>(section) and <b>query</b>(section) into 
something like <a href="baseurl?q=query&s=section">query(section)</a> from the 
output of man -T html -O fragment,man=baseurl where my regex defines query as 
'[a-z._-]+' and section as '[1-9]p?‘. I know that there may be special cases 
where this doesn’t work or where the replacement is triggered falsely. And I’m 
fine with that for my local solution.


> mandoc -Thtml is in very (very) wide use. I think the change you
> described is likely to result in worse markup in some cases.

Thats why I suggested making it an option.


> It would
> be better to encourage developers to put more effort into creating
> high-quality documentation as we do in OpenBSD. And one part of that
> involves picking the right file format to write the documentation in.

Indeed. Fixing the sources would be the best and most correct solution.

I agree that this would require massive support from multiple upstream projects.

For myself I have a working local solution and I’ll live with 
https://man.openbsd.org faithfully displaying the old and broken/semantically 
incomplete formats.


Thanks for your feedback and the observations regarding the root cause. My 
knowledge of mdoc(7), man(7) and other man formats is still rather limited.
-- 
Mike Fischer

Reply via email to