- From: Dave Beckett <dave.beckett@bristol.ac.uk>
- Date: Thu, 7 Apr 2005 16:08:24 +0100
- Cc: public-rdf-dawg@w3.org
I should probably extract some higher level design issues from my review. I found it confusing when the word "pattern" was used alone whether it meant triple patterns, graph patterns (or rarely), query patterns. The definition of graph pattern is scattered and hard to grasp. Related to that, is that value constraints aren't clearly either a graph pattern, part of a graph pattern or part of just one graph pattern (Basic Pattern probably). Several of the formal definitions I found hard to read and instead used the text and examples in preference. It would be good if all the examples were given numbers, anchors and appeared in separate files so they could be tested externally. I put some of the earlier ones in the test suite for this purpose but I don't know if they have been updated to match. I found it rather hard to follow the 10.1 explanation of how to take a variable-binding result (set) and manipulate it. This can include turning it into a sequence and operating on that sequence (ORDER BY, LIMIT, OFFSET) but the process wasn't entirely clear. It also should clearly say - this is for SELECT results only. If CONSTRUCT, DESCRIBE or ASK can use these terms, it probably would be better done in their respective sections. DESCRIBE is looking rather odd, as I learnt it wasn't quite what I thought it was having read through the document fully. I still miss WITH/FROM as it's an obvious omission to not say how to form the data being queried. The expression stuff including datatypes is just working on the detail and looks on the right track, I can't see any big problems with the current approach. Dave
Received on Thursday, 7 April 2005 15:09:42 UTC