- From: <noah_mendelsohn@us.ibm.com>
- Date: Fri, 26 Jul 2002 10:50:58 -0400
- To: "Champion, Mike" <Mike.Champion@SoftwareAG-USA.com>
- Cc: "Christopher B Ferris (E-mail)" <chrisfer@us.ibm.com>, "Dave Hollander (E-mail)" <dmh@contivo.com>, David Fallside <fallside@us.ibm.com>, "Hugo Haas (E-mail)" <hugo@w3.org>, xmlp-comments@w3.org, xmlp-comments-request@w3.org
That you very much for these helpful comments. Not speaking for the protocols WG, but for myself, here are some response on your first two concerns: Your comment #1 ---------------------- I believe comment #1 below represents a misunderstaning. The order in which SOAP headers are examined by a processor for mU faults is NOT determined by the SOAP processing model. Indeed, that's why variability is allowed in the reporting of errors. Headers may be checked in any order. The requirement is that the node must ensure that there will be no mU failures before any other processing of any headers is attempted. Nothing prevents you from doing mU checking in arbitrary order, or for continuing to look for more mU failures if you have found one. The sentence you quote in section 2.4 is: "Therefore, for every mandatory SOAP header block targeted to a node, that node MUST either process the header block or not process the SOAP message at all, and instead generate a fault ..." This does not imply order; it is a declarative statement. I believe that if you follow the rules from 2.6, as summarized in my openning paragraph above, you will find that in the face of mU faults the restriction in 2.4 is automatically satisfied; it will be the case that (presuming some mU failure somewhere in the message), the rules of 2.4 will be satisfied for ALL headers (I.e. a fault will be generated -- note that 2.4 does NOT say a fault specific to that header...it says a fault.) I believe the spec is correct as written. Of course, we can always consider editorial changes if the WG feels that the presentation could be made clearer. I think that it says exactly what it means. Thank you. Your comment #2 ---------------------- You quote: "Tagging SOAP header blocks as mandatory thus assures that such modifications will not be silently (and, presumably, erroneously) ignored by a SOAP node to which the header block is targeted." Note that it says "by the node to which the header block is targeted". You then quote (and comment): "it is not an error for an ultimate SOAP receiver to receive a message containing a mandatory header block that is targeted at a role other than the ones assumed by the ultimate SOAP receiver. This is the case, for example, when a header block has survived erroneously due to a routing or targeting error at a preceding intermediary." That is, it is not an error for mandatory header block to survive erroneously and therefore be silently ignored, thus contradicting the first sentence quoted." No, not a contradiction. In the second case, it is necessarily true that any processing that was skipped was specifically because the message did not reach a node to which the header was targeted. Since it did not reach that node, it could not have been ignored (in the sense of examined and not acted upon) by that node. Again, editorial clarification might be possible, but there is no contradiction in what is intende. The first quote deals with the case where a message reaches its intended target and is not understood; the second points out that the first mechanism is specifically not designed to deal with routing or targeting errors that cause the message to miss its intended target entirely. Again, thank you for taking the trouble to supply these detailed comments. ------------------------------------------------------------------ Noah Mendelsohn Voice: 1-617-693-4036 IBM Corporation Fax: 1-617-693-8676 One Rogers Street Cambridge, MA 02142 ------------------------------------------------------------------ "Champion, Mike" <Mike.Champion@SoftwareAG-USA.com> Sent by: xmlp-comments-request@w3.org 07/26/2002 09:34 AM To: xmlp-comments@w3.org, David Fallside <fallside@us.ibm.com> cc: "Christopher B Ferris (E-mail)" <chrisfer@us.ibm.com>, "Dave Hollander (E-mail)" <dmh@contivo.com>, "Hugo Haas (E-mail)" <hugo@w3.org>, (bcc: Noah Mendelsohn/Cambridge/IBM) Subject: WSAWG Comments on SOAP 1.2 LC WDs David, As requested, and promised by Chris Ferris, here are WSAWG's comments/feedback on the on SOAP 1.2 LC WDs. We have separated out substantive issues or questions from any editorial feedback and request only that the WG respond formally to issues labeled as such (Issue #nn) (e.g. the editorial comments are yours to dispose as the XMLP WG sees fit and require no formal response to WSAWG). Cheers, Mike Champion / Dave Hollander (who thank Chris Ferris for doing the actual work here!) SOAP1.2 Part 1: Messaging Framework Issue #1: Possible inconsistency in SOAP Process Model 2.4 Understanding SOAP Headers, 3rd paragraph, 2nd sentences "Therefore, for every mandatory SOAP header block targeted to a node, that node MUST either process the header block or not process the SOAP message at all, and instead generate a fault ..." This sentence implies that there can be at most one detection of a fault caused by processing a header block as all further message processing ceases on fault detection. However, in various parts of Part 1, there is assumption that multiple misunderstood header blocks may be possibly identified in a fault message, e.g. in 2.6 Processing SOAP Messages, Rule 3: "If one or more of the header blocks identified in the preceding step are not understood by the node ..." e.g. in 5.4.8 SOAP mustUnderstand Faults, 2nd paragraph, 1st sentence: "A SOAP node MAY generate a SOAP fault for any one or more SOAP header blocks ..." Related commentary on this issue: "A SOAP node MAY generate a SOAP fault for any one or more SOAP header blocks that were not understood in a SOAP message." This sentence is very ambiguous and seems to conflict with this sentence from section 2.4 Understanding SOAP Headers: "for every mandatory SOAP header block targeted to a node, that node MUST either process the header block or not process the SOAP message at all, and instead generate a fault" My understanding (according to section 2.4) is that the SOAP node MUST generate a SOAP fault for the first SOAP header block that is targeted at that node, contains mustUnderstand="true", and that it does not understand. When a node encounters a mustUnderstand error, it should stop processing immediately (i.e., no further processing of other header blocks) and return the error. Also, this sentence might be construed to mean that the SOAP node MAY generate multiple faults. If my understanding is correct, then Example 6 should be changed to include only one Misunderstood header blocks. If not, then I would recommend that section 2.4 be changed to indicate that processing may continue after encountering a mustUnderstand fault. Issue #2: Contradiction in SOAP Process Model 2.4 Understanding SOAP Headers, 3rd paragraph, last sentence "Tagging SOAP header blocks as mandatory thus assures that such modifications will not be silently (and, presumably, erroneously) ignored by a SOAP node to which the header block is targeted." This sentence implies that it will not be possible for modifications by mandatory header blocks to be silently and erroneously ignored. However, in the last 2 sentences of the next paragraph, it is said that "it is not an error for an ultimate SOAP receiver to receive a message containing a mandatory header block that is targeted at a role other than the ones assumed by the ultimate SOAP receiver. This is the case, for example, when a header block has survived erroneously due to a routing or targeting error at a preceding intermediary." That is, it is not an error for mandatory header block to survive erroneously and therefore be silently ignored, thus contradicting the first sentence quoted. Issue #3: URI Identification of MEPs Section 3.3 reads[2]: | In general the definition of a message exchange pattern: | * Is named by a URI. | * Describes the life cycle of a message exchange conforming to the | pattern. | * Describes the temporal/causal relationships of multiple messages | exchanged in conformance with the pattern. | * Describes the normal and abnormal termination of a message | exchange conforming to the pattern. This suggests that not all MEPs are identified by a URI. From an architectural point of view, and according to AR009.3[1], all conceptual elements should be addressable directly via a URI. We think that a URI should be required for MEPs: it is easy enough to assign one and will ease identification. Issue #4: External references? Section 2.6 reads[3]: "SOAP nodes MAY make reference to any information in the SOAP envelope when processing a SOAP body or SOAP header block. For example, a caching function can cache the entire SOAP message, if desired." We wonder about external references. They are not explicitely disallowed, but not referenced either. Needless to say, we think that external references are useful, e.g. to reference a large image that one doesn't want to carry along in the message. Issue #5: QName vs URI Sections 5.4.1.3[4] and 5.4.6[5] use and define QNames to identify fault types, whereas WSAWG requirement AR009.3[1] calls for URIs. This is related to two open TAG issues: - Using Qualified Names (QNames) as Identifiers in Content[6]. - Algorithm for creating a URI from a QName[7]. Issue #6 Section 2. SOAP Processing Model. First paragraph. It would be nice to mention the SOAP Response MEP. Also we don't find that section 3.3 SOAP Message Exchange Patterns (MEPs) really provides a description of the relationship between SOAP message exchange patterns and the SOAP extensibility model. It gives only a very brief definition of a MEP, and it makes no explicit mention of the extensibility model. It does describe the relationship between MEPs and the Protocol Binding Framework. Issue #7 Section 2.7.2. Active Intermediaries. Second paragraph: Question. It says, "an active intermediary might have removed and encrypted some or all of the SOAP header blocks found in the inbound message". Shouldn't it say, "some or all of the SOAP header blocks and SOAP body"? Issue #8: Length of URIs. I remember a somewhat long discussion on xml-dist-app about that and am still unsure about what we should say, but the following statement struck me[8]: | SOAP does not place any a priori limit on the length of a URI. Any | SOAP node MUST be able to handle the length of any URI that it | publishes and both SOAP senders and SOAP receivers SHOULD be able to | deal with URIs of at least 2048 characters in length. Even though the URI specification doesn't specify any limit on the length of a URI, the HTTP spec does provide a URI too long error code[9]. 2kB is identified here as a reasonable value to be able to handle. This seems to be an architectural issue, a Web services one as well as a Web one. I am afraid that such a statement will have a big impact on software designed. I think that I would be more comfortable by saying that software should be able to handle URIs of arbitrary length, and that a 2kB-length is seen _in SOAP's context_ as a minimum value to support. Editorial Section 3.3. SOAP Message Exchange Patterns (MEPs) It would be nice to add a reference to Part 2, section 6. Section 4.2. Binding Framework. Third paragraph: ends with "one or more MEP". Should be "one or more MEPs". Section 5.2.1. SOAP header block. In the paragraph following Example 3, in the sentence, "A SOAP sender generating a SOAP message SHOULD use these attributes only on SOAP header block.": It should read, "only on a SOAP header block" Section 5.4.8. SOAP mustUnderstand Faults. First sentence. "When a SOAP node generates a fault with a Value of Code set to "env:MustUnderstand" for," -- delete the word "for". Section 7.3.1. Binding to Application-Specific Protocols. First paragraph, last sentence: "in order to reuse the existing infrastructure associated that protocol" should be "associated with that protocol". SOAP1.2 Part 2: Adjuncts Issue #9: Question Section 6.2.3 State Machine Description, Table 6, Transition Condition, Init row If the Transition Condition is "Unconditional", then the boundary between Init state and the Requesting state is indistinguishable. Shouldn't the two states be one state? Issue #10: Question Section 6.2.3 State Machine Description, last bullet before 6.2.4 "A requesting SOAP node MAY enter the Fail state, and thus abort transmission of the outbound SOAP request, based on information contained in an incoming streamed SOAP response". In this case, should the responding SOAP node switch from generating a SOAP response to a SOAP fault ? Should the requesting SOAP node also generate a SOAP fault since it initated the transmission abortion? Issue #11: Question Section 7.5.2.1 Init, Table 20, SOAP Fault, Malformed Request Message row For Malformed Request Message ( HTTP Status Code 400, HTTP Reason: Bad request) the SOAP Fault is "None". Why not let the SOAP Fault to be "env:Sender", which is consistent with Table 23? Issue #12: Question Section 7.5.2.2 Receiving, Table 23, SOAP Fault to HTTP Status Mapping What about mapping RPC Faults (see 4.4 RPC Faults) to HTTP Status? E.g. rule 2 in 4.4 defines a fault "env:DataEncodingUnknown" which is not mentioned in Table 23. Issue #13: S Section A.1 Rules for mapping application defined names to XML Names, Rule 5, case 4 "Let U1, U2, ... , U8 be the eight hex digits [PROD: 4] such that Ti is "U+" U1 U2 ... U8 in the UCS-4 encoding." The "U+" notation denotes the Unicode code point rather than either UCS-2 or UCS-4 encoding. Moreover, "the full range of Unicode code points [are] from U+0000 to U+10FFFF inclusive; code points above U+10FFFF MUST NOT be used". (See Character Model for the World Wide Web 1.0, 3.5 Reference Processing Model.) Suggest it be change to: "Let U1, U2, ... , U8 be the eight hex digits [PROD: 4] such that Mi is U1 U2 ... U8 in the UCS-4 encoding." Suggest also changing ' "This case implies that Ti has a UCS-2 encoding, which is U+U5U6U7U8." to: "This case implies that Ti has a UCS-2 encoding, which is U5U6U7U8." Issue #14: HTTP binding: interpretation of arbitrary HTTP status codes Section 7.5.1.2 Requesting[8] describes how a SOAP sender reacts to HTTP status codes returned by a SOAP receiver. Table 17[9] only lists a limited number of methods for which next states are listed. This seems to be contrary to the robustness principle cited in Part 1[10]. What would happen if the requesting SOAP node was receiving a 402 Payment Required[11] for example? Editorial Section 5. A Convention for Describing Features and Bindings. First paragraph: "it is used to describe a Request-Response MEP 6.2 Request-Response Message Exchange Pattern and the SOAP HTTP Binding 7. SOAP HTTP Binding elsewhere in this document." It is also used to describe the the 6.4 Web Method Specification Feature feature and the 6.3 SOAP Response Message Exchange Pattern MEP, isn't it? SOAP1.2 Part 0, Primer: Issue #15 2.1. SOAP messages. The paragraph beginning with " A SOAP header". When I first read this paragraph, I was concerned with the implication that SOAP headers should be processed by a SOAP intermediary rather than by a SOAP header processor within a SOAP server. Then it dawned on me that perhaps the authors view a SOAP header processor as a SOAP intermediary. Rather than starting with a description that implies the use of an intermediary, I think the paragraph should start with a basic definition of a SOAP header: A SOAP header is an extension mechanism that provides a way to pass information that isn't application payload. Then it should discuss why you might want to use this information (to pass control, directive, or contextual information) and how to insert header processors into the SOAP processing scheme (either as intermediaries or as header processors called by a SOAP server). Or it might simply make it obvious to the reader that a SOAP header processor is in fact a SOAP intermediary. Issue #16: 2.2.1. Document-style message exchange. Example 2. Shouldn't the returned <p:airportChoices> element contain an enumeration? Issue #17: 2.2.2. Remote procedure calls. This discussion about RPC makes the assumption that it will use the HTTP protocol, and hence spends a great deal of time up front discussing "Web-friendly" concepts before discussing the basics of the the RPC convention. I also find the tone to be somewhat defensive -- It sounds as if it's trying to convince me that this is the right way to do it (and that I have another choice). I much prefer the way this information is presented in Part 2, Section 4 [12]. It's very straight-forward and direct. It says -- this is the way you do RPCs (articulating how it maps to the two MEPs), and when you're using HTTP as a transport, you need to use the Web Method Feature. The GET Web Method is the preferred mechanism to implement SOAP-based idempotent information retrieval. Since the RPC convention isn't dependent on HTTP, I'd recommend: 1) it be explicitly stated right up front (Part 2, section 4 describes it in terms of RPC invocations and responses mapping naturally to HTTP requests and responses, but that RPC is not dependent on HTTP). 2) most of the Web-friendly discussion (The paragraphs starting with "Items 4 and 5" through "We defer until section 3") be moved to section 3 with appropriate references to that section made in this section. Issue #18: 2.2.2. Remote procedure calls. Point #4. Is this "information needed to invoke a SOAP RPC"? Isn't this a clarification of how the information described by points 2 and 3 must be specified? I understand that we want to explain Web-friendly RPCs, but I don't think this point belongs in this list. I think it belongs in the follow-on text with appropriate references to section 3.1.1. I would therefore recommend a change to the following two paragraphs to first assert the requirement to clearly separate the arguments used to identify the target Web resource from the arguments used to convey data or control information and then to explain why. We should promote the Web-friendly way as the preferred way to implement information-retrieval requests, as specified in Section 3.1.3.) Issue #19: 2.2.2. Remote procedure calls. Point #5. This point is only valid when using HTTP. Per Part 2, Section 4, this point should read: "Values for properties as required by any features of the binding to be used. For example, GET or POST for the webmeth:Method property of the 6.4 Web Method Specification Feature." Issue #20: 2.2.2. Remote procedure calls. The paragraph starting with "Items 4 and 5": There should be a reference to HTTP in this paragraph (rather than just referring to the World Wide Web). This paragraph should be moved to section 3. Editorial 1. Introduction. Last sentence of the paragraph starting with "Part 2". I don't think this sentence belongs here. (No other part of this first Introduction section gets into this much detail.) I'm concerned by the implication that SOAP isn't "Web-friendly". At the very least, if the term is introduced here, there must be a definition provided with it. I think the topic is properly covered as part of the introduction in 1.1 Overview section in the paragraph starting with Section 3. 2.2.2. Remote procedure calls. In the paragraph starting with "The central role": This paragraph should be moved to section 3. A space is required here: "Part 2 Section 4.1.1provides" also "data that is necessary" should be "data that are necessary" 2.2.2. Remote procedure calls. Example 4: For readability, here should be a line break between the elements in this line: </m:reservation> <o:creditCard xmlns:o="http://mycompany.example.com/financial"> and this line: </n:name> <o:number>123456789099999</o:number> 2.2.2. Remote procedure calls. Third paragraph from the end, starting with "While in principle": "inependent" is misspelled. 2.4. SOAP processing model. Second paragraph: "(with their contents elided for brevity)" -- we might want to use a more well-known term than "elided" -- perhaps "removed" or "edited". 2.4. SOAP processing model. eleventh paragraph, starting "If a header element": "content is data" should be "contents are data". 2.4. SOAP Processing model. The paragraph starting with "The following table": "appropriate" should be "appropriately". 3. Using various protocol bindings. The paragraph starting with "Thus it is apparent": The phrase "accomplish a particular application semantics" (which appears twice in this paragraph) should be either "accomplish particular application semantics" or "accomplish a particular application semantic". 3. Using various protocol bindings. The paragraph starting with "Among other things": The last sentence reads, "a non-SOAP message acting as a response followed by a SOAP message included as a part of the response" should say, "a non-SOAP message acting as a request followed by a SOAP message included as a part of the response". 3. Using various protocol bindings. The paragraph starting with "Part 2 section 7specifies": A space is required here: "Part 2 section 7specifies" 3.1. HTTP binding. Last sentence of the paragraph starting with "The purpose of providing": "Acceptheader" should be "Accept header" 3.1.1. SOAP HTTP Get usage. The second paragraph, beginning with "Example 8a shows": A space is required here: "URIhttp://travelcompany.example.org/reservations?code=FT35ZBQ" 3.1.1. SOAP HTTP Get usage. The third-to-last paragraph starting with "In Example 8b" "outboundandreturn" shoudl be "outbound and return". 3.1.2. SOAP HTTP POST usage. Example 9: For readability, a line should be inserted between the elements here: </m:reservation> <o:creditCard xmlns:o="http://mycompany.example.com/financial"> and here: </n:name> <o:number>123456789099999</o:number> 3.1.2. SOAP HTTP POST usage. In the paragraph betwee Example 9 and Example 10. "Example 10 shows the RPC return (with details elided)" -- again I recommend using a more common word than "elided". 3.1.2. SOAP HTTP POST usage. Last paragraph. Are we using American or English spelling? Should "behaviour" be "behavior"? 3.1.3. Conveying Web-friendly RPCs. In the paragraph just before Example 12b, starting with "Furthermore": The sentence "SOAP 1.2 recommends that the SOAP Response message exchange pattern be used as described in section 3.1.2." should refer to section 3.1.1. 3.1.3. Conveying Web-friendly RPCs. In the paragraph just before Example 13, starting with "Example 13": "thenameof" should be "the name of" and "Bodyelement" should be "Body element". 3.2. SOAP over email. In the paragraph starting with "Example 15": I recommend using a more commonly known word than "elided". Also a space is required in "email'sMessage-Id". 5. Changes between SOAP 1.1 and SOAP 1.2. Under "Document structure", first bullet: "strucure" should be "structure". References 1. http://www.w3.org/2002/ws/arch/2/06/wd-wsa-reqs-20020605.html#AC009.3 2. http://www.w3.org/TR/2002/WD-soap12-part1-20020626/#soapmep 3. http://www.w3.org/TR/2002/WD-soap12-part1-20020626/#procsoapmsgs 4. http://www.w3.org/TR/2002/WD-soap12-part1-20020626/#faultsubvalueelem 5. http://www.w3.org/TR/2002/WD-soap12-part1-20020626/#faultcodes 6. http://www.w3.org/2001/tag/doc/qnameids.html 7. http://www.w3.org/2001/tag/ilist#rdfmsQnameUriMapping-6 8. http://www.w3.org/TR/2002/WD-soap12-part2-20020626/#http-reqbindwaitstate 9. http://www.w3.org/TR/2002/WD-soap12-part2-20020626/#tabreqstatereqtrans 10. http://www.w3.org/TR/2002/WD-soap12-part1-20020626/#RFC1123 11. http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.4.3 12. http://www.w3.org/TR/soap12-part2/#soapforrpc
Received on Friday, 26 July 2002 10:52:25 UTC