> So maybe we should simply apply the always rules
> to both tables.

This is the path I was going down, but I think you can only safely
"edit" both tables indiscriminately due to the matching component of
the action.

My first pass was adding an internal flag that suppressed the
set/add/merge-like operations when "always" was looking at the
r->headers_out, performing those only on err_headers_out.

But this still leaves unintuitive behavior for things like merge or
append, where the value is at least as likely to start in
r->headers_out, but could be in neither or both tables.  This is why i
finally settled on the idea of changing nothing and trying to document
how 'always' might need to be used to get access to some seemingly
arbitrary pre-existing header values (only when you don't care about
these are the directives mostly intuitive)

-- 
Eric Covener
[email protected]

Reply via email to