Re: Link relation types for non-GET links

[Erik Wilde <[e-mail]>]

hello mike and klaus.

On 2015-08-15 08:27, Klaus Hartke wrote:
> Mike Kelly wrote:
>> You can use a standard HAL link and a link relation if you ensure the link
>> relation is a URL which can be dereferenced to retrieve a machine-readable
>> description of the methods and messages that are possible for that link.
>> Just think of it as machine readable link relation documentation.
> That's an interesting idea. It basically adds a level of indirection,
> which should be useful if many links have the same machine-readable
> description of possible methods and payloads. However, I think I'd
> prefer including that description directly at the link, so a client
> does not have to dereference a URL to understand the link semantics.

if you have a standalone format (such as 
https://github.com/dret/I-D/tree/master/link-desc or whatever you come 
up with to represent link interaction information), then embedding it or 
making it external or allowing both is a simple design choice and flip 
for anybody using it.

our main driver for allowing this kind of flexibility was the scenario 
mentioned before: the general link relation type might define a number 
of parameters clients can submit, but the specific runtime info could be 
more constrained ("only 42 pages you can GET"), and clients could adjust 
accordingly. ideally, there would be some kind of "inheritance model" 
(effective link interaction semantics are a combination of the generic 
ones, and runtime info can selectively overwrite), but we never got far 
enough to really have a well-defined model for that.

cheers,

dret.

-- 
erik wilde | mailto:[e-mail]  -  tel:[phone] |
            | UC Berkeley  -  School of Information (ISchool) |
            | http://dret.net/netdret http://twitter.com/dret |