Thanks Frederick and Marcos - responses inline. Only a couple of questions left :)
Regards, Mark -----Original Message----- From: marcosscace...@gmail.com [mailto:marcosscace...@gmail.com] On Behalf Of Marcos Caceres Sent: 22 April 2009 11:46 To: Frederick Hirsch; Priestley, Mark, VF-Group Cc: Barstow Art (Nokia-CIC/Boston); public-webapps Subject: Re: [widgets] New WD of Widgets 1.0: Digital Signatures spec published on March 31 On Tue, Apr 21, 2009 at 11:14 PM, Frederick Hirsch <frederick.hir...@nokia.com> wrote: > Mark > > Please find responses inline. Thanks for the review. > > regards, Frederick > > Frederick Hirsch > Nokia > > > > On Apr 7, 2009, at 2:27 AM, ext Priestley, Mark, VF-Group wrote: > >> >> Hi Art, All, >> >> Please find below my editorial comments and requests for >> clarifications based on the new WD [1]. While it is a long list the >> comments are all minor and so hopefully easily addressed. Overall I >> think the spec is looking good, for which a lot of thanks must go to >> Frederick and Marcos! >> >> That said, I have a couple of more substantive comments that I will >> send in the next couple of days. >> >> Many thanks, >> >> Mark >> >> >> [1] http://www.w3.org/TR/2009/WD-widgets-digsig-20090331/ >> >> ----- >> COMMENTS >> ----- >> >> 1.0 >> >> "A widget package can be signed by the author of the widget producing >> an [XMLDSIG11] signature that cryptographically includes all of the >> file entries other than signature files. A widget package can also be >> signed by one or more distributors, with XML signatures that each >> cryptographically includes all of the non-signature file entries as >> well as any author signature." >> >> Change the last sentence for consistency between definitions, ie: >> >> "... A widget package can also be signed by one or more distributors >> <change> of the widget, producing [XMLDSIG11] </change> signatures >> that each cryptographically includes all of the non-signature file >> entries as well as any author signature." > > ok [mp] Thanks > >> >> >> ----- >> Can we remove the following sentence? This is a general property of >> signatures which I'm not sure we need to include. >> >> "Digitally signing implies use of private key material only known by >> the signer, thus enabling verification of integrity and signature source." > > ok [mp] Thanks > >> ----- >> 1.1 >> >> We don't actually define any XML elements in the >> "http://www.w3.org/ns/widgets-digsig"; namespace... is this worth >> noting this and/or removing the "wsig" prefix? >> > > We define URIs using this namespace so we should keep the URI definition. > ok with removing prefix since it is not used now but would prefer to > keep to avoid errors later. Not a big issue to remove though. [mp] I'm OK either way. > >> ----- >> The terms "XML elements" and "resources" seem to be used >> interchangeably? Is there a difference? > > yes, one is xml elements others are resources as referenced by URI Mark, I'm worried you asked this question? Is there confusion somewhere wrt to the use resource and xml elements? [mp] No, it's mostly a case of me needing to read the text more carefully! My confusion was caused by the fact we only define the namespace prefixes that we use in throughout the spec in the context of resources. >> >> >> ----- >> "Algorithms used by XML Security are defined in a number of >> places..." - could we tighten up this sentence, eg something like >> "This specification references algorithms defined in [XMLSecAlgs] and >> [XMLDSIG11]" ? >> > > No, XMLSecAlgs does not define the algs. There are defined in a number > of places :) OK - my concern was just that [XMLSecAlgs] cross references lots of algorithms that we don't need but happy to leave as it is. > >> ----- >> 1.2 >> >> "compressed (or Stored)" - either remove capitalisation of Stored or >> add it to compressed >> > > > I suggest "stored". Marcos? Stored should probably be [Stored], with a reference to the RFC for the algorithm. [mp] OK for me >> ----- >> "physical file" -> file ? >> > > Marcos? ok with file personally Agree. [mp] Thanks >> ----- >> "top-most path level" - is there a better way of saying this? >> > > don't think so unless you have a proposal without using the word "root" I know it's nasty, but people understand it. Lets play wordsmith only once we have all the tech stuff solved. [mp] As I can't think of anything better, happy to leave as is. >> ----- >> "which MAY logically contain" - if the configuration file is made >> mandatory then the MAY should be a MUST >> > > I think it is a MAY, others? Technically, Mark is correct. But leave it as a MAY (or maybe drop MAY altogether) because this spec does not put conformance criteria on packaging. [mp] proposal "of the widget package, which logically contain one or more file entries, as defined" Note that reading this again - if a file entry is a file or a folder, then there must be at least one file entry unless the widget is an empty package (and if it's signed it can't be empty because at a minimum there would be a signature file entry!) so I think one one or more is correct. > >> ----- >> Question: is a file entry the same as a file? If so then we should >> always use "file entry" in place of "file". If not then perhaps we >> should define file? >> > > I don't think they are the same. This is a P&C question. Marcos? Depends. A file entry is the representation of file data in a zip archive. A file is a physical uncompressed file as would appear on disk. If we assume that signatures will act on physical files, it will be correct to talk about "files". [mp] OK, makes sense. >> ----- >> "(i.e., a third party that is distributing the widget on behalf of >> the author)." - in some cases the author may also be (one of) the >> distributor(s). suggest changing "i.e." to "e.g." >> > > I think i.e. is correct. In the case you suggest they just happen to > be the same entity fulfilling two roles. [mp] OK > >> ----- >> 3 >> >> "As well as sections marked as non-normative, examples and notes in >> this specification are non-normative. Everything else in this >> specification is normative. The security considerations section is >> non-normative." >> >> Suggest change to the following for readability: >> >> "As well as sections marked as non-normative, the examples and notes, >> and security considerations in this specification are non-normative. >> Everything else in this specification is normative." > > yes, better. [mp] thanks > > >> >> >> ----- >> 4 >> >> "This section defines how to locate digital signatures in a widget >> package for processing. An implementation MUST achieve the same >> result as the following algorithm used to locate digital signatures >> in a widget package:" >> >> In the sentence above, change "digital signatures" to "signature files" >> (in both cases) >> > > ok [mp] Thanks > >> ----- >> "This MAY be determined by the security policy used by the user agent." >> >> Can we say will or, better yet, SHOULD or MUST? Otherwise, what do we >> mean when we say MAY? Who (what) else may enforce security policy? > > we mean may since security policy is out of scope. How can we mandate > what is not defined? [mp] Well if we think of security policy in the broadest sense, any decision of which signatures to process is part of the security policy. The UA has to do something! Changing "used" to "enforced" could make it clearer. Alternatively simply deleting the sentence would work. > >> >> >> ----- >> "Thus the highest numbered distributor signature would be validated >> first." >> >> Change to: >> >> "Thus in the case that one or more distributor signatures were >> validated, the highest numbered distributor signature would be >> validated first." > > ok [mp] Thanks > >> >> >> ----- >> 5.1 >> >> "A widget package MAY be digitally signed using XML Signature >> [XMLDSIG11]." >> >> don't we mean: >> >> "A widget package MAY be digitally signed using the profile of XML >> Signature [XMLDSIG11] defined by this specification." ? >> > > ok [mp] Thanks > >> ----- >> As this section is talking about generating a signature, I suggest >> that we remove "and validated" in the following sentence: >> >> "Each signature file MUST be generated and validated in" >> > No - 6.1 applies to both generation and validation, common to both. [mp] You're right - my mistake > >> ----- >> 5.2 >> >> As per previous email exchange we need to re-work author signature >> definition > >> ----- >> "zero or one author signatures." - remove final "s" > > No, I think that the current is correct grammatical usage and clear in > meaning. [mp] OK > >> >> >> ----- >> "This represents the digital signature of the author of the widget >> package." >> >> add "signature file" ie "This signature file represents the digital >> signature of the author of the widget package." >> > ok [mp] Thanks > >> ----- >> 5.3 >> >> "This represents the digital signature of a distributor of the widget >> package." >> >> add "signature file" ie "This signature file represents the digital >> signature of a distributor of the widget package." >> > > ok [mp] Thanks > >> ----- >> 5.3.1 >> >> "Within a widget package these signature files MUST be ordered based >> on the numeric portion of the signature file name. >> >> Thus, for example, signature2.xml precedes signature11.xml." >> >> Question: what does this mean? What is it requiring from a widget >> package? >> >> ----- >> 5.4 >> >> "Implementations MUST be prepared to accept X.509 v3 certificates >> [RFC5280]." >> >> Can we say "User agents" rather than implementations >> > we mean implementations [mp] OK - but we earlier say "A user agent is an implementation that attempts to support this specification." Therefore isn't a UA also an implementation? > >> ----- >> 5.5 >> >> "It MUST be unique for a given signer." >> >> Do we need to make it clear that we are not expecting the UA to check >> this? I take it we're not asking the UA to check this, right? > > What do you suggest we say? [mp] Proposal: "Signing parties are expected to ensure that the dsp:Identifier signature property value is unique for the widgets that they sign" > > >> >> >> ----- >> 7.1 >> >> "Each ds:Signature property" -> "Each ds:SignatureProperty" ? >> > > meant as written since wanted to be clear about properties as opposed > to XML representation. [mp] OK > >> ----- >> In step 5 there is no bullet for digest algorithms, which there >> probably should be. >> > > I believe digest algorithms are mentioned for ds:References for the > digesting of content, but not needed in step 5 since the signature > method includes the digest method (eg RSAwithSHA256) [mp] OK - digest algorithms are covered in 3.b. > >> ----- >> 7.2 >> >> "This MUST be a unique signing string for all signature files created >> by the signer." - same comment as 5.5. ie - Do we need to make it >> clear that we are not expecting the UA to check this? > > What do you suggest we say? [mp] Proposal (as above): "Signing parties are expected to ensure that the dsp:Identifier signature property value is unique for the widgets that they sign" > >> >> >> ----- >> 7.3 >> >> "If signature file validation fails for any reason, any external >> entities (e.g., a user agent that implements [Widgets Packaging]) >> relying on the validation of the signature file MUST be notified of >> the failure..." >> >> Maybe we should say that notification of successful validation must >> also be provided? > > Add before the last paragraph?: > > If signature validation is successful any external entities (e.g., a > user agent that implements [Widgets Packaging]) relying on the > validation of the signature file MUST be notified of the success. [mp] Perfect - thanks > >> >> >> ----- >> 8 >> >> "A signature file may also be renamed, which can affect processing." >> suggest modification to "...which can affect the order in which >> distributor signatures are processed" > > ok [mp] Thanks > >> >> >> ----- >> 9.1.1 >> >> "Upon signature generation, if this property is used, the value is >> set to ..." >> >> Is inconsistent with the sentence from 5.1 which states: >> >> "Each signature file MUST contain a dsp:Identifier signature >> properties element compliant with XML Signature Properties >> [XMLDSIG-Properties] and this specification." >> > > this is not inconsistent. Section 9 says if used, section 5.1 says it > is used in the profile... > >> Suggest deletion of ", if this property is used," from the first >> sentence > > I do not think I understand the rationale for this change. [mp] It was meant to remove the inconsistency, however, re-reading the intro to section 9 (which I managed to skip even though it's in bold, red text) I realise that this section is to be removed from the Widget 1.0: Digital Signatures specification, so there is no inconsistency. My mistake. > >> >> >> ----- >> 9.1.2 >> >> "Profiles MUST specify details of the identifier property value >> creation and interpretation." What does "Profiles" mean in this >> sentence > > the widgets signature specification is a profile... > [mp] See above comment - my mistake >> >> >> ----- >> "If multiple instances of this property are found on a single >> signature, then applications MUST NOT deem any of these properties >> valid." - which would in turn mean that the signature was invalid, >> right? We may want to state this? > > the properties are not valid though the signature still might be valid. > Interpretation of properties is profile dependent. [mp] and again, based on a misunderstanding of why this text was included. > >> >> >> ----- >> 9.2 >> >> Note that the same comments may apply to 9.2.1 and 9.2.2 dependent on >> the discussions on the mandatory/optional status of this property. > > same answers as for 9.1.2 > [mp] and same response - my error >> >> >> >>> -----Original Message----- >>> From: public-webapps-requ...@w3.org >>> [mailto:public-webapps-requ...@w3.org] On Behalf Of Arthur Barstow >>> Sent: 02 April 2009 17:21 >>> To: public-webapps >>> Subject: [widgets] New WD of Widgets 1.0: Digital Signatures spec >>> published on March 31 >>> >>> On March 31 a new WD of the Widgets 1.0 Digital Signature spec was >>> published and announced on the W3C's home page: >>> >>> [[ >>> 2009-03-31: The Web Applications Working Group has published a >>> Working Draft of Widgets 1.0: Digital Signatures. This document >>> defines a profile of the XML Signature Syntax and Processing 1.1 >>> specification to allow a widget package to be digitally signed. >>> Widget authors and distributors can digitally sign widgets as a >>> trust and quality assurance mechanism. Prior to instantiation, a >>> user agent can use the digital signature to verify the integrity of >>> the widget package and perform source authentication. This document >>> specifies conformance requirements on both widget packages and user >>> agents. >>> ]] >>> >>> Please review this new WD as soon as possible, preferably within the >>> next two weeks: >>> >>> <http://www.w3.org/TR/2009/WD-widgets-digsig-20090331/> >>> >>> -Regards, Art Barstow >>> >>> >> > > > -- Marcos Caceres http://datadriven.com.au
