The necessary flexibility might be achieved if the text read: "The value of the text response entity body MAY be determined by running the following illustrative algorithm:". My point is that the algorithms are illustrative, devised specifically for clarity (as is pointed out in the document), yet the text also demands that you MUST use the algorithms as outlined. This is a logical conflict in the use of the RFC 2119 terminology. I noted this conflict when I opened the document and did what many technical people would do and go directly to the technical illustrations. There I saw that the algorithms were mandated. It was only (after this initial disappointment) when I read the document completely that I noted that the preamble told a different story. I am merely pointing out a logical inconsistency in a Normative section of the document. One might characterise it as a typo. But as this is a Working Draft, there is the chance that this could be amended so that anyone extracting just the Normative portion that refers to the algorithm(s) will also include text that appropriately highlights that the algorithm is essentially illustrative, though conformant to the functional behaviour demanded by the specification. Regarding your comment commencing "Yes, they are not required to do so.", I fully agree. This was not the point of my comment. I completely understand that different implementations are desireable. In fact, this is why I was initially disappointed when I went directly to the algorithms and found that the text indicated that I MUST implement them in the way they were written. That, of course, is not the message you had intended to convey. This is a minor issue. I have no intention of making any further comment or clarification, and will let the WebAPI people decide for themselves if there is any merit in it. Regards, ---Rotan.
________________________________ From: Anne van Kesteren [mailto:[EMAIL PROTECTED] Sent: Tue 19/06/2007 20:26 To: Rotan Hanrahan; [email protected] Subject: Re: Status of algorithms On Tue, 19 Jun 2007 18:11:25 +0200, Rotan Hanrahan <[EMAIL PROTECTED]> wrote: > I am concerned by subtle conflicts in the normative text. For example, > you read the following: > > "The value of the text response entity body MUST be determined by > running the following algorithm:" > > The use of MUST as per RFC 2119 in this text makes the algorithm > Normative, while the text you quote says: "User agents MAY optimize any > algorithm..." Yes, they are not required to do so. What it basically says is that user agents are free to implement the specification in any way they want as long as they return the same results as the specification would for all possible inputs. We can't even always tell what algorithms they have used to implement the functionality. For instance, with closed source implementations. The statement is there just to avoid confusion. (The sentence is in fact taken from the XBL 2.0 specification as suggested by someone during the Last Call phase.) > I fully accept that it is intended that optimizations will be permitted. > It's just a formulation of the text that caused a little confusion for > me. The normative text: "The value of the text response entity body MUST > be determined by running the following algorithm," doesn't give the > necessary flexibility that is intended. Could you give an example that shows why this is not the case? -- Anne van Kesteren <http://annevankesteren.nl/> <http://www.opera.com/>
