Re: Composability

Norman Walsh wrote:
> Three things motivate me to attempt to solve this problem in V1:
> 
> 1. If you encounter the problem, there's no straightforward workaround.

Right. You have to solve it the same way you solve it in XSLT: you 
generate the pipeline based on the XPath, and then run the pipeline. 
This would be fiddly but doable if we had a standard run-pipeline step, 
but we decided against that (for good reasons). As it is, it requires a 
"manual" (or automatic, just not with XProc) two-stage process.

> 2. The real screw case is where you attempt to use a prefix that the
>    pipeline author used, but with a different binding. In that case
>    you don't get an error, you just get the wrong result. Languages
>    shouldn't allow that sort of user error to go undetected.
> 
> 3. I'm an engineer.
> 
> On the other hand, the longer it takes for us to get V1 out, and the
> more complicated we make it, the less likely it is to be successful.
> So there are pragmatic reasons to say, "not in V1".

Yes, I know, and I've said several times that I didn't think there was 
the will or time to pursue this. But you did keep pushing...

Of course, ideally I'd like to fix it, but I'd be more-or-less happy for 
a v1 that has:

  (a) a clear story about what *does* get passed into invoked steps by 
way of namespace bindings and variable/option bindings.

  (b) a *BIG* warning of the dangers of passing XPaths or QNames around 
as option values

  (c) a relatively clear upgrade path to a v2 that provides a fix!

And as long as we can start working on a v2 pretty quickly.

cheers,

Jeni
-- 
Jeni Tennison
http://www.jenitennison.com

Received on Friday, 8 June 2007 20:12:55 UTC