I think the cases where the change would make a functional difference
or impact performance would be quite rare, unless you commonly have
processors or pipelines that leave _all_ their declared outputs
hanging.

Also, the idea is not to blindly run _every_ processor. For example,
only the p:when or p:otherwise block that is chosen would be
"executed".

There is clearly more thinking to do around this :-)

-Erik

Damon Rand wrote:
> Hi there,
>     +1 for leaving the XPL spec the way it is now. ;-)
>
> Using a null-serializer to force a processor to be read is good
> because then the programmer explicitly controls the exact behavior
> of the pipeline. Isn't there also a situation where a <choose> can
> decide whether to read the output of one processor or another at
> runtime? In this situation you certainly wouldn't want a processor
> to be executed even if it is never read.

>>We had a fair amount of debate around this, and maybe it would make
>>sense to give processors that declare outputs which are not
>>connected a chance to execute as well. Because of the policy of lazy
>>evaluation of pipelines, this wouldn't necessarily solve, for
>>example, the case of an XSLT transformer which is not connected. It
>>would be up to each processor implementation to determine whether it
>>has a side effect and to execute or not. I think this would make
>>sense.

_______________________________________________
oxf-users mailing list
[EMAIL PROTECTED]
http://mail.orbeon.com/mailman/listinfo/oxf-users

Reply via email to