- From: Glen Daniels <gdaniels@sonicsoftware.com>
- Date: Mon, 9 Jan 2006 15:42:13 -0500
- To: "David Illsley" <david.illsley@uk.ibm.com>
- Cc: <paul.downey@bt.com>, <public-ws-addressing-tests@w3.org>
Hi David, all: > I asked the exact same question about why not require this > last year. The answer surprised me (probably shouldn't have), > but I can see the argument. > If we mandate that the testid is in the body a less than > honest 'implementor' could simply grep an incoming message > and return a correct one. Big deal - you can still do a simple XSLT or regexp match to return a correct response too! In fact, if you have the correct set of XSLTs, who's to say you *don't* have a valid WSA implementation? :) (remember Don Box's XSLT SOAP implementation?) Besides, I'm not suggesting that the strings be entirely static here; but simply that there be SOME way in the message to indicate what test conditions are supposedly being satisfied. Really, why is this even controversial? > I'd like to agree on a format for us to use for debugging > purposes though the implementations should support random > strings (with "" causing a fault) Here's the "benefit" I see in supporting truly random strings (or rather, in failing to have a standard test identifier somewhere in the message): you lose the ability for automated validation of test results. Gee, that's awesome. :) Can someone point me to another benefit? --Glen
Received on Monday, 9 January 2006 20:42:44 UTC