Re: LDP-Server - Issue-57

hello arnaud.

On 2013-06-14 8:12 , "Arnaud Le Hors" <lehors@us.ibm.com> wrote:
>>Do you think that we have here one of those differences that goes beyond
>> language? That we have here a notion so subtle that there remains no
>> way to tell the difference in language itself?
>I think the difference really lies into whether you see the interaction
>model and the RDF type as inseparable or not.

very well put. to spice things up a little (sorry arnaud :-)... no,
seriously, please don't respond to this (i promise i won't), but maybe use
it as a little thought experiment for yourself:

we theoretically should be able to "re-encode" LDP in JSON, saying clearly
which hypermedia affordances clients really need to know about, and then
we encode those in JSON. the payload of the JSON-guided interactions then
is RDF (because that's what we're shipping back and forth), but the
service is driven by LDP interactions that can be represented any way we
choose.

for RDF-minded clients, of course that would be inconvenient and thus we
very don't want this. it's pretty much always a bad idea to mix metamodels
in a service, it makes processing hard (you need two processing
machineries) and is more complex than serializing everything in the same
way.

but as arnaud rightly pointed out: if somebody took off and said "i love
LDP, but i really want to build a JSON flavor of it", they should be able
to fairly easily extract the interaction model, represent it in a way that
makes JSON-LDP consumers happy, and then wrap that set of interactions
around the RDF that's being exchanged.

so, please *do not* respond ;-) but i think arnaud made an excellent
comment here, pointing out a very helpful separation of concerns (service
model vs. data model).

cheers,

dret.

Received on Friday, 14 June 2013 15:32:38 UTC