Re: ldp-ISSUE-98 (bertails): HTTP status code for application specific errors [Linked Data Platform Spec]

On Thu, Apr 24, 2014 at 2:26 PM, Linked Data Platform (LDP) Working Group
Issue Tracker <sysbot+tracker@w3.org> wrote:

> ldp-ISSUE-98 (bertails): HTTP status code for application specific errors
> [Linked Data Platform Spec]
>
> http://www.w3.org/2012/ldp/track/issues/98
>
> Raised by: Alexandre Bertails
> On product: Linked Data Platform Spec
>
> LDP does not defined a recommended status code for application specific
> errors. 4.2.1.6 [1] currently says that the status code is in the 4xx
> range, plus the use of the rel=describedby Link header.
>
> Here is the discussion I had this morning with SteveS:
>
> [[
> <betehess> SteveS, looks like the LDP spec is vague re: status code for
> business logic errors (I guess on purpose :-). It says 4xx +
> rel=describedby.  Unofficially, which 4xx would you choose?
> <SteveS> 400
> <betehess> my colleagues think that 400 looks too much like a fallback
> strategy
> <SteveS> I agree with the sentiment that 400 is “fall back” but I don’t
> see anything better
> <SteveS> perhaps we should create/propose one?
> <betehess> that would be a good idea
> <betehess> I would expect the spec to help me choose one for business
> logic errors
> <betehess> but that would be kinda a fallback as well
> <betehess> so maybe 400 is just good enough when used with rel=describedby
> <betehess> or we need something targeting specifically application
> specific logic
> <SteveS> yes, well what we have now is the link and you follow it and
> maybe what you get helps
> <SteveS> we could improve it by a) a different status code and/or b) a
> specialized link relation
> ]]
>
> I am totally fine with the 3 following solutions:
>
> 1. 400 + rel=describedby
> 2. new status code
> 3. 400 + specialized link relation
>
> I have a slight preference for 1. as it is less work.
>
> Alexandre
>
> [1] http://www.w3.org/TR/ldp/#h5_ldpr-gen-pubclireqs
>
>
I'm for #1 and putting #2 and/or #3 on wish list.  I think once we have
"shapes" a solution may be more obvious.


- Steve Speicher