[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: EDXL-DE 2.0 for the F2F - Objectivity, Subjectivity andInterpretation.
Agreed. Actually after I wrote the unsigned int comment, I realized that the spec doesn’t state that you can’t have a positive floating point value (which would need to have a range specified) to indicate a fraction of a km. From: Gilmore, Timothy [mailto:TIMOTHY.D.GILMORE@saic.com] Don, As far as your question about using different wording about OPTIONAL, MAY use multiple… (and other related issues) (From our Engineer) RFC 2119 defines these words, and they are in common use in other RFCs and various standards documents. These words do *not* by themselves indicate subjectivity. If the spec says "this element is OPTIONAL," then the element is optional. A message containing the element conforms. A message not containing the element conforms. A mechanical system can make that determination. Every test engineer will arrive at the same conclusion. Similar logic holds for phrases like "MAY use multiple." Take the <digest> element inside EDXL-DE. According to the spec, it's an xsd:string, which is zero or more "printable" characters, such as "A," "/," "$," etc. According to the spec it calculated using the Secure Hash Algorithm SHA-1. SHA-1 produces a 160-bit output. Right off the top of my head, I can come up with half a dozen ways to encode those 160 bits into an xsd:string, most of which could be "the most obvious way." What did OASIS intend? Did OASIS intend something else? What if a vendor comes up with a way we didn't think of? Does the message conform? No mechanical solution can accommodate that vendor; different test engineers will arrive at different conclusions. Take the <combinedConfidentiality> element inside EDXL-DE. The spec says REQUIRED, MUST be used once and only once. There's not much subjectivity there, is there? Then again, the spec also says "...the most restrictive of the <confidentiality> elements..." So if one <confidentiality> element contains "I hope no one finds out," another contains "don't tell my sister," and a third one contains "φος" (the Greek word for "light"), which one is the most restrictive? It comes down to the question of every engineer coming to the same conclusion: If there's room for interpretation, then it's subjective. Even if there are multiple choices, but no wiggle room within any of the choices, the decision is objective. Consider the <mimeType> element. An infinite number of strings are MIME type, but any given string either is a MIME type or it's not. My reply to Donald would be that those words are fine. Additionally, however, I would note that his proposed statement about radii is ambiguous until a "normal signed or unsigned int." is defined. 25 years ago, a "normal unsigned int" was 16 bits wide. 15 years ago, it was 32 bits wide. Today, it is 64 bits wide. Unless you've been working on CDC mainframes for 30 years, in which case it has always been 30 or 36 bits wide. Coincidentally, a unsigned 16-bit int with a value less than the maximum value of a signed 16-bit int isn't too far off: that would limit radii to 32767 kilometers, which is about 80% of the Earth's circumference. Timothy D. Gilmore | SAIC Sr. Test Engineer | ILPSG | NIMS Support Center | IPAWS CA / NIMS STEP phone: 606.274.2063 | fax: 606.274.2025 mobile: 606.219.7882 | email: P Please consider the environment before printing this email. From: McGarry, Donald P. [mailto:dmcgarry@mitre.org] Tim- I wholeheartedly agree! I did bring this up for discussion earlier and we agreed that a circle should be <circle>lat’,’lon<space>radius</circle> Which makes comment 1 and the example wrong (extra space in both between the lat and lon). This is on the issues list for 2.0. I will add the point about the radius, because as stated it should be an unsigned integer with a maximum value less than that of a normal signed or unsigned int. Are you suggesting that we use different wording for the OPTIONAL, MAY use multiple? That was a little confusing to me at first, so input would be appreciated. I have added these topics to the issues list From: Gilmore, Timothy [mailto:TIMOTHY.D.GILMORE@saic.com] All, Some of the things we look at are objectivity and subjectivity due to our accreditation under the American Association for Laboratory Accreditation (A2LA) for NIMS STEP and IPAWS Conformity Assessment (CA) testing. Many elements under the OASIS EDXL suite of standards including CAP use words such as “SHOULD” and “MAY” which are clearly subjective in nature. One of our engineers pointed out some issues that we should keep in mind when going over the EDXL-DE 2.0 document during the F2F. For CAP: What we're looking for are rules or constraints that are open to interpretation, or not fully specified, rather than being completely "nailed down." For example, consider the <circle> element. Is the following a "correct" <circle> element? <circle> 0, 0, 150000000 </circle> It certainly fits the descriptions in that element's comments: (1) it's in the form "latitude, longitude, radius"; (2) the central point conforms to WSG84; (3) the radius value is expressed in kilometers; and (4) it is a properly escaped XML string. Then again, the radius of the circle is approximately the distance between the Earth and the Sun. Note that the given definition includes the word "geographic" (twice!) and that the center of the circle is specified as longitude and latitude, all of which indicates to me that the circle ought be to Earth-bound. Someone else may interpret the standard differently, and the standard doesn't put a real limit on the radius of the circle. The point is that the standard doesn't really specify enough for a tester to determine whether or not a <circle> element is conforming. The tester has to make up his (or her!) own rules to complete the test. Multiple testers will certainly come to different conclusions, and all will be correct to within the subjectivity allowed by the standard. (And that all said, note that the given example doesn't match the form given in comment 1; the comma between the longitude and the radius is missing. Since all of section 3 of this standard is normative, this is a bug in this standard.) For another example, consider the <senderRole> element. The standard says "OPTIONAL, MAY use multiple." Despite the words "OPTIONAL" and "MAY," an individual tester can determine without a doubt whether a given message contains zero or more <senderRole> elements, and an infinite number of testers (all else being equal) will come to exactly the same conclusion. Perhaps something to think about at the F2F. Thanks, Timothy D. Gilmore | SAIC Sr. Test Engineer | ILPSG | NIMS Support Center | IPAWS CA / NIMS STEP phone: 606.274.2063 | fax: 606.274.2025 mobile: 606.219.7882 | email: P Please consider the environment before printing this email. |
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]