So, reading this thread carefully, one possible resolution is that the
atom "rel" attribute is extended.

<link rel="parent" href="http://myblog/monthly_summary">
<link rel="replyto" href="http://yourblog/contentious_article">
<link rel="reference" href="http://communityblog/about_this_topic">

These 3 to be added to the exisiting: related, alternate, vias,
enclosure and self.

Aleksander Slominski wrote:

> if you have more complicated links relations i think it is beneficial
> to define superset of atom:link and allow more powerful relations in
> decentralized way - i did it internally in my engine by using QNames,
> for example to point to internal topic (which is like "tag" or Wiki
> topic): <a href="/SomeTopic" rel="s:topic">...</a> that is later
> translated to <s:link rel="s:topic" xmlns:s="my-space"
> href="SomeTopic"/> where s:link is superset of atom:link with
> redefined rel/rev and other attributes possible.
>
>
.. and Aleksander appears to agree. Is there any mechanism to tell the
aggregator about the superset relationship (RDF would say superclass)
between s:link and atom:link ?

> and for interoperability i can still of course generate <atom:link> in
> my atom feed (alas! i loose my custom relations as atom 1.0 has
> centralized rel content and does not allow namespace

So what's the best, to get Atom buy in (which may not happen), or to
roll your own namespace (which will doubtless get ignored by the
aggregators)?
I've more or less taken the latter route but using RDF properties from
known vocabularies (eg Annotea 'annotates'
http://www.w3.org/2000/10/annotation-ns).

Steve