Comments on CMIS link relations

[Mark Nottingham <mnot@mnot.net>]

[ CC:ing atom-syntax FYI ]

CMIS <http://www.oasis-open.org/committees/download.php/32171/Draft%2061.zip 
 > specifies a large number of new link relations for use in Atom. A  
few comments and questions follow:

1. Each one of the link relations specifies a type of document that it  
references (with "Mime/Type", although I note that the proper term is  
media type, and the values given are prose descriptions, not media  
types). Is the intent here to limit these relations to those types? If  
so, this is conflating the job of a link relation with a media type.  
Link relation types should not be specific to any single format.

2. Some of the proposed registrations seem to overlap with existing  
relation types (e.g., "parents" whereas "up" has already been  
registered; "repository" where "service" would probably do.).

3. Other proposed registrations seem to be very specific to your use  
case (e.g., "streams", "allowableactions"). These cases may be better  
served by using extension relations (i.e., URIs).

4. Of the remaining ones, it does seem like there are some useful  
things to register (e.g., "child", "latestversion"), but the language  
shouldn't be specific to your use case; they need to be generic.

5. In case you're not aware, there's a proposal circulating to revise  
the link relation registration process, as well as provide a framework  
for them; see <https://datatracker.ietf.org/drafts/draft-nottingham-http-link-header/ 
 >.

Cheers,

--
Mark Nottingham     http://www.mnot.net/