[Eis-devel] What to do with errors...

[Bas L. <ba...@ba...>]

Hi All,

I have been working on the servermanager application and got the feeling 
that the way user errors are currently handled in EIS is not
the best way to go. So please think with me, and let me know how you 
think that errors should be handled.

My first question is whether a service should stop immediately after an 
error is found. EIS1.5 services do not and try to find all errors
and return them as an array. The other option is to stop when the first 
error is found and return only one. One error is more simple than 
finding all errors and will provide one specific reason why a process is 
cancelled with a single error message. An array of errors can be 
helpfull when many things are wrong that can be fixed by the user before 
resubmitting.

The second question is at what point translation of error messages 
should be handled. One option is to do translation in the service when 
an error is detected. This is easy because all the information for 
construction is directly available but also means that when a service is 
invoked it will have to be language aware. This is a bit awkward when 
calling services via xml-rpc for example. Another option is to leave 
translation of error messages to elements or rpc clients. This is a more 
presentation independent solution but does mean that the client will 
need to be able to translate string identifiers to text and that 
information that has to be replaced in the messages will have to be 
provided as well. For example, a message like "The username 'johndoe' is 
already in use" should be constructed from a code "e.username.taken" and 
the replacement "johndoe". This would mean that error objects become 
more complex than just numbers or strings.

My current preference would be to stop when the first error is found and 
to do translation in elements or rpc-clients.
What are your thoughts?

greetz,
Bas