Technical issues for F2F

[sbrown@metastorm.com]

Aplogies for late posting, and again if any of these are duplicates - I'm a little behind in my reading. Here are 3 issues I'd like to get aired:

XPath: the schema allows a queryLanguage to be specified for a process, defaulting to XPath 1.0. Fine - but elesewhere in the doc it seems to state consistently that XPath 1.0 is used: shouldn't all these references to XPath 1.0 be qualified with "(or whatever queryLanguage is specified for the process)"?

Flow semantics: Section 5 (p14) states the general principle that "compliant implementations MAY choose to perform static analysis to detect and reject process definitions that may have undefined sematics." In contrast, Section 12.5 (p64) states "a link MUST NOT create a control cycle... Therefore, directed graphs created by links are always acyclic." Why not relax this restriction and allow cyclic graphs? I understand the issues with parallelism and cyclic behavior, but it's nevertheless possible and often useful to define a semantically 'safe' cyclic graph - an example being a simple state transition diagram or flowchart.

Exception handling wrt super-(and sub-)sets of BPEL: Allt he implementations I've seen to date have extended BPEL, and some may well implement subsets. Should there not be a well-defined behavior for a BPEL engine (or compiler) that encounters an extension it doesn't understand? (maybe there is and I've missed it). And should there be a minimal subset that any implementation MUST implement to get the BPEL sticker?

Thanks,
Steve