See also: IRC log
With corrections from ER and NM.
<Norm> Scribe: Norman Walsh
<Norm> ScribeNick: Norm
Date: 03 Jan 2006
<DanC> hmmm... I have ht's "What is a namespace, anyway?" message flagged for response... would be nice to have one or more issues connected to it for prioritization
I would guess it's related to nsState-48 and perhaps the whole "grounded in the web" issue of self-describing documents that ht and I still have open.
Hey, timbl, I sent you something about that before Christmas, did you ever get a chance to read it?
<DanC> I think you'd rather I *didn't* associate it with nsSate-48, right, ndw?
<DanC> Vincent, I see actions re issue 8 still in http://www.w3.org/2001/tag/2005/03/action-summary.html and not synced with the issues list
<ht> Well, I have to confess that email (http://lists.w3.org/Archives/Public/www-tag/2005Dec/0120.html) did start from our discussion of nsState-48 last time . . .
<noah_montreal> +n5c2 n6ah
They looked fine to me
ER: They looked fine to me too
RESOLUTION: Approved
No regrets given
<DanC> minutes 20 Dec (1.2 2006/01/03 18:05:15)
RESOLUTION: Confirmed; DO to scribe, ER in his absence
DC: Related actions pointers to go the issues
list
... As far as I can tell, the old pending list still has novel information
VQ: I've made progress on the issues list but
the actions are not yet up-to-date
... The only complete action list we have is still the separate small
list.
... I'm still planning to move everything to the issues list.
RESOLUTION: Agenda accepted
NDW proposed http://lists.w3.org/Archives/Public/www-tag/2006Jan/0007.html
<DanC> looks good... [[ An XML namespace has a namespace name (a URI) and a set of local
<DanC> names (NCNames as defined in [XML Namespaces]). ]]
NDW: That's what I came up with from the minutes
HT: I can live with it, but I'd like to see if
we could live with more.
... I wondered if we have consensus about what the namespace name
identifies.
... Sometimes I think it identifies a set of names and sometimes I think it
identifies a namespace. Since we don't have a good definition of the latter
,that's not helpful
... Suppose I said: "A namespace is identified by a namespace URI (aka the
namespace name)" Would that attract consensus.
DC: It's tautologically true, but not useful
<Zakim> timbl, you wanted to say that there is a pun going on: the ns uri is a string and a URI of a document
HT: I'd like to find an answer to the question "what is identified by a namespace uri" in the webarch document.
TBL: The namespace URI identifies a namespace
document. The namespace doesn't have a URI; it's a set of names which start
with this common prefix which is kind of a string.
... It's a little architectural kludge. It happens to be the same URI used
for all the names, but it identifies the namespace document.
... You could have a separate URI for the namespace, e.g.,
namespace-document-uri#thisnames, but it's not really worth doing
... Maybe we should be able to talk about namespaces in the abstract, but we
don't very often.
NM: I
<Zakim> noah, you wanted to remind ourselves that we should decide whether we're talking specifically about namespaces in XML
NM: I've been troubled about our lack of clarity about when we're talking about namespaces in XML (a W3C Rec) vs. namespaces in a broader sense which almost certainly include namespaces as used in RDF in the abstract which are often serializeable in XML and could go on to include anything on the web that feels like a structured namespace.
<noah> [Definition: An XML namespace is identified by an IRI reference; element and attribute names may be placed in an XML namespace using the mechanisms described in this specification. ]
NM: I understood the history of this issue to
be largely about Namespaces in XML.
... I'm happier just to avoid scope creep and keep this finding focused on
namespaces in XML.
... This issue started by raising the question "gee, we've got these XML
namespaces (specifically the xml: one), and some folks think they're mutable
and some don't and we need to say something about that"
... I think we've all agreed with NDW's analysis. I don't think we should go
very far in restating or bending what the recommendations say about what the
URI identifies or anything else.
<Zakim> DanC, you wanted to say that I prefer not to dereference (use) namespace names; I prefer to combine them with localnames to come up with a URI-term and look that up and to note
DC: If you're happy with NDW's text, then we're fine. If you're saying namespaces in RDF are different than namespaces in XML, then I object.
NM: I don't think we need to go there for purposes of this issue.
<ht> HST wonders why the httpRange-14 compromize isn't the right way to approach the identifies a namespace/a namespace document issue . . .
TBL: The TAG document says that this URI
identifies the namespace document. Now we've got this recommendation that
says "identifies this namespace". One way out is to say that it's indirect
identification.
... You can say that the namespace is the one that has that namespace
document, for example.
... We can weaken the sense of "identifies" in the namespace document.
<noah> I agree with Tim that there is some issue as to whether the URI should identify the document, the namespaces, or perhaps the document as representative of the namespace. What I'm quesitonning is whether we need to go into any of that in order to resolve >this< issue.
<ht> "namespace URI can be used to identify an information resource that contains useful information,"
<Zakim> noah, you wanted to say HTTP range 14 doesn't talk about documents, it talks about info resources
DC/TBL discuss where this is actually grounded in the webarch document
<ht> http://www.w3.org/TR/webarch/#namespace-document
NM: The httpRange-14 resolution talks about info resources not documents
TBL: Do you really want to go there?
NM: My definition of an information resource leaves open the possibility that our httpRange-14 resolution allows me to return 200 for the namespace itself. I think they fit well in a computer message.
TBL: Informally, we mean "document" when we say information resource
NM: My understanding is that TBL takes the view that "information resource" is either a synonym of document or very close
TBL: Yes
VQ: I'd like to separate nsDocument-8 and
namespaceState-48. Can we reach a conclusion about 48?
... We can let the discussion about namespace documents go to the next
item.
<timbl> I am happy with the text.
VQ: To me the last issue that was relevant to
the finding was just this first paragraph. NDW has rewritten it, I have heard
several people agreeing with (or satisified with) what NDW has drafted. Can
we focus on that paragraph?
... Does anyone have a problem with it?
<Roy> link?
DC: I'm happy with that paragraph but I have another comment
<noah> No problem with Norm's proposed text.
VQ: Is there consensus on NDW's para?
... Yes, I conclude that there is.
DC: The good practice suggests that it's OK not to have it in the namespace document. That seems bad.
NDW points two paragraphs done.
DC: Can we combine the two shoulds into one.
<noah2> +1 to separating them
HT: No that's weaker because it says if you don't have a namespace document you're off the hook.
<Roy> s/namespace name and a local name, the qualified/namespace name and a local name: the qualified/?
DC: Yes, you're off the hook but in the doghouse.
NM: I agree with HT and NDW.
<EdR> I like it the way it is.
VQ: I guess we can keep the document as it is for that part. Ok, DC?
DC: Yeah.
VQ: Any other comments?
NDW makes the change RF suggested
<DanC> (ndw, can you save to http://www.w3.org/2001/tag/doc/namespaceState.html )
<EdR> +1 to approve now
NDW: Proposed: approve NDW's namespaceState-48 finding with the new paragraph and the editorial suggestions proposed by RF and NM as an approved TAG finding.
<DanC> 2nd
<Roy> +1
VQ: Anyone object?
RESOLUTION: the proposal carries
<scribe> ACTION: make the changes, publish the finding, and post to www-tag [recorded in http://www.w3.org/2006/01/03-tagmem-minutes.html#action01]
DC: Who's waiting for this?
NDW: Core was waiting, but since xml:id has already gone to REC...
NDW promises to dot the i's and cross the t's wrt XML Core after he's made the announcement
DC: I wonder if there's anyone else that cares
DC asks about xml:base
NDW pushes back successfully.
NDW: The problem with mentioning xml:base is simply that it would cause more editorial changes than I'm comfortable making after we've approved the finding :-)
<noah> scribenick: noah
NW: I noticed last time that the examples were
bad and that the appendices were confusing people.
... The 13th Dec. draft more carefully uses the actual RDDL namespaces for
natures and purposes.
... I've also added the list of RDDL natures and purposes to sections 5 &
6. BTW: those sections are therefore no longer blank.
Noah also thanks Norm for adopting some of his more minor suggestions.
VQ: Who else has reviewed this?
NM: I've reviewed the lastest and I'm quite happy with it.
RF: I've skimmed.
<Norm> RF: there's not really an example of a namespace that isn't flat in nature
<scribe> scribenick: Norm
RF: I suggest mentioning a namespace where the namespace isn't flat
DC: Clearly HTML is in that space. Constructing a namespace document for HTML would be much harder.
RF: I don't have a better example in mind
DO: Do you mean something like WSDL that has symbol spaces?
RF: I was thinking of XML documents where there'd be a link element named "a" and some other element with an attribute named "a"'
TBL: Isn't that what you meant?
DO: Yes
HT: I can't think of any that are as well know as HTML that have an example of this
DC: I think it'd be great if it wasn't quite so well known.
TBL: Should we as the TAG say that this is a bug. These things aren't universal names.
<Roy> and also the case where the element (e.g., <address>) means different things depending on what elements encapsulate it.
DO: How are they not universal if the fragid specification tells you how to make the names?
<Zakim> noah, you wanted to suggest that deprecating symbol-space-like constructs is its own issue, not a no brainer
TBL gives the example of "cite" in HTML meaning either the element or the attribute
NM: I don't think that this example is
sufficiently obvious or straightforward that we want to slip it into this
finding. Maybe this is a good new issue.
... I think there's a lot to discuss before we get there.
DC: I don't think you need to split the issue.
NM outlines why he thinks its a different issue
NM: I think your critique is about the nature of the namespace not the namespace document
<ht> Well, the schema for schema documents has two distinct element types whose local name is 'group', and that name also is used for an identity constraint
DC: I think a lot of folks won't be a happy if we close this issue without addressing this issue
TBL: we should point out that there are cases where the names aren't universal and maybe point to another issue.
NM: That's what I was saying. It's too big to go in a namespace document
<Zakim> Roy, you wanted to suggest adding an algorithmic mapping of names to URIs within the RDDL document
<noah> Actually, I said it's beyond the scope of this >issue<
DC: I'd be happy just ot not close any of these issues until we close the "self describing documents" issue.
<ht> I would have said 'not unique' or 'not unequivocal' rather than 'not universal'. . .
<noah> as this issue is about Namespace Documents, not the design of namespaces themselves.
RF: I'd suggest that this is one of those places where we can address the question independently of various opinions about what's an appropriate namespace. A lot of these things already exist. I'd rather have a way to say "if you need to know the URI of something" then here's how to do it.
<noah> Therefore, I would prefer to open a new TAG issue on whether symbol spaces are or are not bad practice. I don't think we have consensus to say that here, and I think that in any case doing so would be beyond the scope of this issue.
RF: I'd rather point to a namesapce document
that describes an algorithm for doing the mapping
... For example, in a flat namespace it could be concat(uri,'name'). There
are lots of ways (editorially) that this could be expressed.
TBL: For a more complex example, I guess we could point to the WSDL case.
<Zakim> ht, you wanted to ask Roy about expressing sorts in URIs via path components vs. via parameters
<ht> i.e. .../1999/xhtml/element#cite vs. 1999/xhtml/?sort=element&name=cite
HT: Suppose we take one of these example, Roy, do you have an inclination about putting the "sort" in the path or the parameter or a syntactically complex fragid.
RF: It's irrelevant to me so what's important is that there be no restriction.
VQ: It's still not clear if we need a new issue
<DanC> (when TimBL says "cite from HTML" is ambiguous, one possibility is: no, it's not ambiguous; it refers to the element. attributes don't have top-level names. Another is to say html#cite is indeed a hosed URI; it's ambiguous; don't use it. use html#element_cite or html#attribute_cite)
DO: I think it should be in this finding. It seems awfully darned related.
<ht> DanC, right -- that's the third (syntactically complex fragid) approach
<noah> My main concern is that we not try to do a rush job on symbol spaces. If the group wants to take the time to do a careful analysis and see whether it fits well here, I have no objection.
NDW proposes to make a stab at expanding section 4
<Roy> +1
NDW agrees to have it done by 17 Jan
<ht> DanC, none of the three approaches in their simplest form will cope with arbitrarily nested scopes, as in W3C XML Schema (and many programming languages)
<Roy> NDW's proposal
<noah> Fine with me too.
DC: I'd rather not try to patch this. I don't want to prempt the stuff on self describing documents
DC expresses concern about the number of hours in the day and the fact that NDW is on the hook for self-describing documents
<Roy> http://www.w3.org/2001/tag/doc/nsDocuments/#div.fragid
TBL: Are you proposing that we widen the scope
DC: Yes, let's talk about the whole range of
issues related to self-describing documents right now
... That's the highest priority for me at this point
NM ponders how we'll structure our time if we do this
DC: I don't want to add an issue on
self-describing documents because it's already in so many of our existing
issues. We keep dividing ht equestion so we can't actually talk about what we
need to talk about.
... I think the best bang for the buck is to discuss something on shared
text.
<noah> I'm surprised. I think self describing documents makes a great standalone finding.
DC: I think ns-8 is responsive as it is.
<noah> Me too. That's one reason I didn't want to broaden now, as we are close to publishing something useful.
HT: Agrees assuming we add something to the preface about the complexity of non-uniqueness.
NDW ponders calling this fiding finished and doing the same work he earlier proposed under a different title
RF: If the topic comes up again in the future, I think section 4 is where the answer belongs.
VQ asks DC about closing it
<ht> HST would like the docbook example graph diagram to use 'validation' as the purpose of the top two links
DC: I'm of several minds. It's acceptable to me to try to call it done as it is.
<DanC> <http://www.w3.org/2005/12/assoc#>
<noah> Should the doc state that this URI is new for purposes of this doc?
<DanC> 404 there
DC: I want to point out that 2005/12/assoc# is created by this document. The 404 is unacceptable.
<Roy> It is acceptable to me to call it done for now, but I think the TAG should consider adding such an algorithm statement (how to contruct URI for each term in my namespace) to NS descriptions and then revising section 4 accordingly
<DanC> (ouch. still finding typos in the examples.)
In draft finding: assoc:relaxng-validation should be purpose:relaxng-validation
HT: The purposes of both these first two is
validation.
... In the prose you say "validation"
NM: I think this leads to a long discussion in
which both sides are right.
... I can imagine use cases where it really matters that they provide
overlapping but different services and some use cases where they really are
the same
HT: At the very least they need to be made consistent.
VQ: I hear consensus about publishing the
document more-or-less as it is, just fixing a few details.
... Is there any objection about that?
DC: And closing issue 8?
VQ: Yes, and closing issue 8.
HT: Why is the ontology non-normative.
NDW professes ignorance in ontology creation
HT agrees it's tricky.
DC: Now we're into issue RDF meaning...
<DanC> rdfURIMeaning-39
NDW: It's the prose that's normative. The ontology is just an expression.
DC: asks about following your nose
<Zakim> DanC, you wanted to recall why we can't close 8 just yet
<DanC> "DanC to ask for "default nature" to be changed to "implicit nature" in RDDL spec"
HT: agrees you don't find these statements now, but I believe Jonathan Borden would add the link if the community agrees
DC: I thin it'd be nice if App B said "we'd like the RDDL folks to point to this"
HT: I wouldn't mind asking him first. This has been pointed to from www-tag several times.
DC: I have an action earlier, I could add this to that.
<Zakim> timbl, you wanted to asl whether these nature URIs identify natures as URIs.
TBL: Asks about the nature of natures
<timbl> http://www.ietf.org/rfc/rfc2026.txt
<timbl> http://www.iso.ch/
TBL: Is "http://www.ietf.org/rfc/rfc2026.txt"
the nature of an IETF RFC?
... Is "http://www.iso.ch/" the nature of an
ISO spec?
DC: In RDDL land, yes. Pretty weird, but yes.
TBL: So these aren't URIs?
DC: How do you mean?
<timbl> http://www.ietf.org/rfc/rfc2026.txt
TBL: Because ...rfc2026.txt is the URI of an
RFC so how can that be a nature?
... I don't think that's the architecture they're using.
HT: I think it is fairly parallel to the discussion we had earlier about indirect identification.
DC: So the rddl.org stuff has some waffly prose and such but what this finding says is that if you GRDDL it you can believe the RDF. Whatever wooliness was in the spec, you can now appeal to RDF semantics. The RFC *is* the nature.
HT: net net: good or bad?
<DanC> assoc:nature
DC: Borderline acceptable.
... We own 'assoc:nature' so we're saying that the nature of an RFC is an
RFC.
HT: That one's a bit of a pun. But take the nature of HTML 4.
TBL: In all of these where there isn't a hash,
I don't like the architecture.
... In one case it's a document defining a language, in another case it's the
home page of an organization.
<DanC> (yes, I was wondering if timbl understood what this finding says. I'm glad he's swapping it in, though it does seem to be undoing the proposal we almost resolved.)
TBL expresses concern about the fact that he can't conclude anything from the fact that something is a nature
HT explains how the standard use case works.
TBL: I think these things are squatting in URI space, they aren't really URIs. You can't use the ISO home page without asking their permission.
<timbl> natureRelatedSomehowTo
NDW: I can see how it's weird but these URIs are already deployed.
DC: We could change the mapping. We could create a more complex mapping.
<timbl> DanC: We could make the RDF say 'the nature is something whioch is an org wih this homepage:"
<Zakim> noah, you wanted to talk about layering
NDW expresses concern about making the model more complex because it's going to make getting community consensus more difficult
NM: There's a yin-yang thing here, where
sometimes a very carefully layered ontology is constructed and sometimes
"rough and ready" weirdness "just works". But at least if I know what I want
to put in a RDDL document it's the same thing I put in the top of my schema
document.
... If for each of these 18 things I have to go somewhere to figure out how
to change it, that makes the problem even worse.
... Given that it's deployed and there are operational advantages to just
leaving it alone.
... The same URI is being used for the nature and the beast itself. It's not
ideal, but I'm inclined to leave the RDDL world alone.
<DanC> (but we're not leaving the rddl world alone. we're coining assoc:nature )
NM: Each person doing a new one does it their own way.
TBL: I feel the other way. I'd never want to
use any of the RDF statements from these RDDL documents in my system.
... There may be a large number of people who feel that way now, but I'm not
willing to put a huge spoke in the reusability of data for this.
... I think we should change them all.
NM: You're impling that by using the URI for ISO here you'd be encouraging people to confuse ISO with a nature. But I'd have thought that what comes out of here is an RDF predicate. If NDW defines this not as "this is a nature" but "this is a resource that reminds you of a nature" I'd have thought this would be ok.
<DanC> interesting possibility... assoc:nature is "a resource that will remind you of the nature"
Scribe fails to capture the point TBL tries to make
TBL: objects to "remind" because it's hard to see how that would be understandable to a machine.
NM: Proposes a mechanical transformation
DC: How does that help
<ht> I still like an anonymous node whose xxx:namedByRDDL1.0With property is http://www.iso.ch/
NM: Formally on the RDF side whenever you are refering to the nature you get a URI that isn't the URI of ISO
TBL: proposes how you might turn all the URIs in to local names.
<timbl> http://www.rddl.org/natures/
VQ: At this point we're running out of time.
TBL: We're in the middle of a conversation.
VQ: I propose that we get back to this next week but to try to decide how to progress. We need to figure out how to organize the work for the future.
<DanC> (tbl, do you consider your action re http://www.w3.org/1999/10/nsuri to be relevant to this issue, 8? hmm... perhaps more relevant to nsState48)
VQ: Think about how we can make progress and
we'll come back to this next week.
... We also need to talk about the last call documents from the CDF next
week.
ADJOURNED