Re: Base URL handling

>      <location xml:base="http://download.example.com/root"
> href="packages/example-0.1-1.x86_64.rpm"/>
> 
> My reading of RFC 3986 suggests that the combined URL should be:
> 
> http://download.example.com/packages/example-0.1-1.x86_64.rpm

After reading the RFC, I must agree.

   o  return a string consisting of the reference's path component
      appended to all but the last segment of the base URI's path (i.e.,
      excluding any characters after the right-most "/" in the base URI
      path, or excluding the entire base URI path if it does not contain
      any "/" characters).

That's how urlparse.urljoin() works.

> However, the expected URL (from the server directory layout and the
> fact that yum can download the RPM) appears to be:
> 
> http://download.example.com/root/packages/example-0.1-1.x86_64.rpm

Yes, instead of merging, Yum just joins all components, inserting
a slash in the middle if there's not one already.

> Is there a high-level description of how yum combines URLs?  Any
> suggestions for approximating that?

Yum handles base URLs in variety of ways:

1) metalink.xml

The <url> elements always end with "/repodata/repomd.xml".  Yum checks
for this and strips these last two path components.  This is probably
an accepted de-facto standard.

2) baseurl set in *.repo file

If the baseurl does not end with "/", Yum includes the last component
in full url.

3) repomd.xml

The <location base=".."> is parsed, but never used.  Mirror URLs are used
instead.

4) primary.xml

That's the example you used.  Handled as in #2, with one small extra bug,
that "/" may be duplicated.  createrepo sets the baseurl attribute directly
from the --baseurl cli option.

IMO, #1 can't be changed.  #2 is mostly ok, as the shipped .repo files
almost always have baseurls with trailing slashes anyway.

I'm surprised that baseurls in #3 and #4 are handled differently.

#4 could be worked around in createrepo, by appending a trailing slash
to --baseurl option.