- From: Jeni Tennison <jeni@jenitennison.com>
- Date: Thu, 27 Mar 2008 15:37:10 +0000
- To: public-xml-processing-model-wg@w3.org
Norman Walsh wrote: > / Alessandro Vernet <avernet@orbeon.com> was heard to say: > | On Mar 26, 2008, at 7:31 AM, Henry S. Thompson wrote: > |>> I don't think that's a necessary restriction. There's no reason why I > |>> shouldn't be able to initialize two or three variables right before > |>> the p:choose that needs them. > |> > |> Hmm. I guess. Given that I think we definitely should _evaluate_ > |> them all at the beginning, I thought it was simpler to _put_ them all > |> at the beginning. . . > | > | I don't think we need to say that variables are all evaluated at the > | beginning. Instead, each variable can be evaluated the first time it > | is used. It may render the implementation slightly more complicated, > | but not unreasonably so. > > Interestingly, I reached the opposite conclusion thinking about it > this morning. > > The document order of steps is irrelevant in practice. That means we'd > rules about what happens to variables that occur between steps when > the steps are reordered. I don't think it'd be easy to understand > those rules. Couldn't you say that the bindings visible to any steps within a given subpipeline are all the variables that are bound within that subpipeline, and leave it for the implementation to decide whether it evaluates all the variables before running the steps, or evaluates them on demand, or what? If the order doesn't matter (to the implementation) then we ought to let people put them in whatever order feels right to them (just as we have for steps). Jeni -- Jeni Tennison http://www.jenitennison.com/
Received on Thursday, 27 March 2008 15:37:45 UTC