[OAI-implementers] error questions

[Simeon Warner]

On Mon, 24 Jun 2002, Paul Cummins wrote:
>    I know the protocol states that a single error is enough to trigger
> an error response but it seems to imply that all codes should be
> returned.
>    If in the case of the illegal argument, where a value like
> test=zzzzzz is sent, should this short circuit the testing process 
> and jump straight to the error response without testing from, until,
> set, etc., since someone could be trying to break something?

I think it is allowable to either short circuit checks or to test other
arguments.
 
>   Should a multiple argument error be "badArgument" and then should we
> indicate which word was repeated?  (example below)
>
>    repository/oai?verb=Identify&set=set1&from=2000-01-01&set=set2
> 
> errors:
> <error code="badArgument">The query contained an illegal
> argument--set</error>
> <error code="badArgument">The query contained an illegal
> argument--from</error>
> <error code="badArgument">The query contained an illegal
> argument--set</error>
> 
>   Are these all of the codes or should there be another one because
> Identify should have no other parameters?

I think the response above is fine.

The philosophy should be to provide error messages that are helpful
to someone trying to work out why their harvester isn't working.

--
Simeon



PS. My server checks for duplicate arguments first because the 
arguments are then stored in a hash. Thus the error messages 
from the request above are:

<error code="badArgument">Found illegal duplicate of argument 'set'.</error>
<error code="badArgument">Verb 'Identify', illegal argument 'set'.</error>
<error code="badArgument">Verb 'Identify', illegal argument 'from'.</error>