Re: [httpapi] linkset: use of "absolute URIs", corner cases, formatting

[Erik Wilde <[e-mail]>]

hello christian.

On 2021-06-15 10:17, Christian Amsüss wrote:
> Thanks, the section of 8288 is pretty explicit -- so it may just be
> worth pointing out that such links can multiply in size when expressed
> in linkset. (Think of what `</>;rel="a b c d e f g h i j k l m n o p
> q";foo="hello multiplication factor"` would look like).

you're right about this example, but this scenario looks so extremely 
unlikely to happen for web links that i don't think we need to cover it 
as a specific example.

> Plus, a
> back-converter should be advised to aggregate links that only differ in
> their rel.

that's up to them, as far as i am concerned. maybe i would do this when 
implementing one, but that's really up to personal preferences, as both 
ways are semantically equivalent.

> Maybe then it's more practical to describe what *is* round-tripped (that
> is, the information can be round-tripped, but serialization details
> including order can't).

do you think this should be mentioned explicitly? if so, please let us 
know in an issue and possibly point out where you think this information 
should be included. thanks!

> I fail to see the difference; in
> </foo>;rel=hasMoreInfo;type=application/linkset, the type says that
> thiis is a set of web links, and the hasMoreInfo strongly implies (no
> less strongly thatn rel=linkset would) that the link context is
> mentioned in there.

i am not quite following you. hasMoreInfo is not a type that exists. 
we're about web linking. we want to be able to link to sets of links. 
those can be linksets as defined by us, or by other specs (such as the 
CoRE spec).

link relations are about linking because you expect a certain affordance 
(as in this case: i want to find more links about that resource). the 
media type (if provided) then says how this represented. it may not be 
provided, or could be content-negotiated. pushing everything into the 
media type (and therefore requiring it and making the link relation type 
essentially pointless) to me seems like an anti-pattern for web linking.

> If that further discussion is on record anywhere, I'm happy to read up
> to understand the difference point, but just from Section 3 I don't, and
> still see rel=linkset to nail down two orthogonal aspects of a link
> relation where one aspect is already covered by another target
> attribute.

https://datatracker.ietf.org/doc/html/rfc8288#section-2.1 is the best i 
can provide here. there is a history of separating the "what" to find 
aspect (link relation type) from the "how" aspect of how it is 
represented (media type). we simply follow that general pattern.

thanks and cheers,

dret.

-- 
erik wilde | mailto:[e-mail] |
            | http://dret.net/netdret    |
            | http://twitter.com/dret    |