- From: Kay, Michael <Michael.Kay@softwareag.com>
- Date: Fri, 20 Jun 2003 19:28:07 +0200
- To: Jim Fuller <jim.fuller@stuartlawrence.com>, public-qt-comments@w3.org
A personal response from me: you will get an official response when the WG has considered your comments. However, I would suggest that if you want the WG to consider your suggestions seriously, you need to provide more detailed semantics for the proposed functions, and rationale for their inclusion. > fn:node-kind > some sugar: what about cdata type test? can we add a > whitespace node type test? What would a "cdata type test" do? We don't have CDATA nodes in the data model. What would a "whitespace node type test" do? This would seem to be a test on the content of the node, not on its type. You can test whether the string value of a node is all whitespace by using [normalize-space($node) eq ""]. > > why not add a lightweight schematron-like fn:assert() > function, probably nearthe the fn:trace section > fn:assert(test,pass element path, fail element path) > ex.fn:assert(count(//fingers) = 4, > system/message/pass_element, system/message/fail_element) > note that this could be used in xacml type situations as well I can't see how this would be any different from an if/then/else test, with a call to error() in one of the branches? Generally we are trying to reduce the size of the specifications rather than add in more "nice to have" features. > > as fn:trace is currently defined, why not have fn:log > equivelent function, might help some non-english speakers How would you see the functionality of fn:log differing from fn:trace? It is difficult to guess the semantics of your proposed function from its name alone. > > why not add fn:random as per latest exslt rand:random() > debate; integer * ranged-random-sequence(integer minValue, > integer maxValue[,integer numberOfItems := 1 [,double seed := > (double)current-time()]]) There are many, many functions that we could consider adding, for example trigonometrical and square-root functions. We have to apply the test "does this need to be in the core, or could it be provided in a third party library?". Generally, we are trying to keep the size of the core library small, and many people (both within the working group and outside) feel it is much too big already. Michael Kay
Received on Friday, 20 June 2003 13:28:33 UTC