On 13/12/12 16:45, Francesco Chicchiriccò wrote:
On 13/12/2012 14:27, Sergey Beryozkin wrote:
Hi
On 13/12/12 12:42, Andrei Shakirin wrote:
Hi,
I am working on Rest interface migration (to Apache CXF) and
analysing exception propagation from service to client.
Actually SyncopeClientCompositeErrorException is sent using:
a) ExceptionType HTTP header containing exception type (enumeration
SyncopeClientExceptionType)
b)<ExceptionType>.element HTTP header as list of strings for error
details (that are used to fill SyncopeClientException.elements)
I find that fine at the moment, as far as details information is only
simple and short list of strings. Potentially it is possible to have
more complex and long structures in exception details. Therefore the
question is does it make sense to use HTTP header only for
ExceptionType and send details in HTTP body?
It means that we will change network protocol between client and
service, not sure how critical it is for existing Syncope users.
There are 2 alternatives:
a) leave the propagation as it is: send type and details in HTTP
headers. In the future additional information exceptional still can
be sent with HTTP body.
b) send only ExceptionType in HTTP header and details element in HTTP
body.
Do you have any preferences for (a) and (b) or there are other
alternatives?
Perhaps it might make sense to keep a simple HTTP header anyway, for
the receivers to be optionally able to quickly check the exception type
without having to read the response, and also return the body with all
the details, including the actual exception type, so that the whole
info can then be analyzed on the client side after the request has
been completed, so one more possible option :-)
Hi,
I like this (c) option as well:
1. ExceptionType + elements in the HTTP header (like as now)
2. HTML (short) body with more verbose description, possibly localized
(1) could be used by 3rd party REST clients while (2) could be used to
provide better error message for admin console.
Hi, I'm rereading and I guess I may've repeated what Andrei already
suggested with his option a) :-)
Either way, looks like we all in agreement
Cheers, Sergey
Regards.
