[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [xri] Grammatical rules for cross-reference usage
|
What I meant by “identical from the standpoint of
resolution” is not that *example and !example are equivalent, but that
resolution behaves no differently. In other words, both are just subsegments,
and there is no difference in the XRI resolution process between a *subsegment
and a !subsegment. However from a “grammar” standpoint they “mean”
very different things about the identifier. =Drummond From: Schleiff, Marty
[mailto:marty.schleiff@boeing.com] I don't understand how *example and
!example are absolutely identical from the standpoint of resolution. When I
suggested we consider them to be equivalent (with the ! being meerly a claim of
persistence that can be ignored for purposes of resolution and equivalence
checking) you over-ruled me by insisting they are in two different
namespaces. If they're in two different namespaces, I don't see how they can be
identical from the standpoint of resolution. Marty.Schleiff@boeing.com; CISSP
From: Drummond
Reed [mailto:drummond.reed@cordance.net] Les, I know you have been saying, “if
there is a grammatical difference then it should mean that we do something
different in the only client we define, resolution”. After some
reflection, I no longer believe that’s true. Here’s my logic: the XRIs *example
and !example are absolutely identical from the standpoint of resolution.
However “grammatically” they have very different meanings –
one is used for a reassignable identifier and one is use for a persistent
identifier. That’s very important to XRI authors and XRI consumers, but
not to the resolution infrastructure. Therefore, the test I think we should be
applying is the one Marty has been asking for: “If there is a grammatical
difference, XRI authors should know when to use which type of cross-reference
– direct or parenthetical.” I believe the answer posted today on http://wiki.oasis-open.org/xri/XriCd02/DirectConcatenation
is crystal clear: use a direct cross-reference when you are only crossing two
contexts, and use a parenthetical cross-reference when you must cross three
contexts. The first will be very common, the second will be rare – as
rare as putting English words in “quotes” or (parentheticals). The fact that we started with the rare
case before we discovered the common case is, IMHO, what has caused our
deliberations to take so long. It’s not without precedent – 2.5
years ago the TC struggled for several months over switching from dots and
semicolons to stars and bangs as subsegment delimiters. Today we take it for
granted that you can use dots in a i-name. I know I sound like a broken record,
but I believe the same will be true for direct concatenation. I agree with you that it’s great to
finally be getting input from others on the TC who have not been as close to
the discussion. I also look forward to the feedback from non-TC members that
we’ll be getting at the XRI sessions we’ll be holding at IIW next
week. I too can hardly wait for us to close and
get XRI Syntax 2.1 in production. I’m going to be concentrating hard
tomorrow and Sunday on posting the first editor’s draft of XRI Resolution
2.0 Working Draft 11 so we can finally get that into production. =Drummond From: Chasen, Les
[mailto:les.chasen@neustar.biz] I am not getting the grammatical
difference either. We have discussed many different grammars over the
past few weeks. There was the difference between a word in quotes and not
in quotes, there was talking about “THE” or “A” entity
in the xref, there was some discussion of direct concatenation meta data being
factual real data about the entity in the xref and in the local context being
the parents opinionated view on the xref, and now possession for direct
concatenation and I’m not sure I understand the parenthetical option. I have been saying for a long time that
IMHO if there is a grammatical difference then it should mean that we do
something different in the only client we define, resolution. If there is
no significant difference and therefore no difference in behavior then we are
back to equivalence. We have also discussed a potential middle
ground of allowing direct concatenation of the two GCS characters that are
currently not resolvable, + and $. The editors have been debating this issue
for some time. I think we have posted a lot of good info to this list
about the issues. I would like to strongly encourage the other members of
this list to please speak up. I understand the wish to have a simpler syntax
one that is more intuitive to end users. I am all for it but it
doesn’t come about by magic. To me it is up to the community on how
we proceed. From: Schleiff, Marty
[mailto:marty.schleiff@boeing.com] Hi Drummond (& All), I'm somewhat relieved that in both cases
(without parens and with parens) the referenced global XRI is an identifier for
the target resource that can be used independently of the parent context. So,
"@bbb" ALWAYS means the same named resource, whether it stands alone,
or is in an xref with parens like "@aaa+(@bbb)", or is referenced in
direct concatenation like "@aaa@bbb". I think the grammatical implications of
each (without parents and with parens) are insufficiently described.
You've described a syntactic difference that has no meaningful explanation. Why
would someone care to cross two contexts as opposed to crossing three
contexts? You are still searching for meaningful grammer to
assign to the syntax, instead of designing syntax to support meaningful
grammar. This is backwards, akin to deriving the abstract syntax from the
concrete syntax. Marty.Schleiff@boeing.com;
CISSP From: Drummond
Reed [mailto:drummond.reed@cordance.net] As promised, I just posted an update to
the Direct Concatenation wiki page with the “grammatical rules” for
cross-reference usage. See section 4 of:
http://wiki.oasis-open.org/xri/XriCd02/DirectConcatenation
I believe – and I know I’m
going way out on a limb here – that they will satisfy *both* Marty and Les. The reason is that
they address Marty’s concern that the rules proposed last week were
counterintuitive because they required a parenthetical cross-reference to NOT represent
a globally-identified resource, and also Les’s concern that the
difference between direct concatenation and parenthetical encapsulation must be
clean and clear. And the very best part? They are simple
and straightforward with no funky exceptions. Feedback gladly solicited. =Drummond From: Drummond Reed
[mailto:drummond.reed@cordance.net] I find these arguments persuasive.
I’m working this afternoon on addressing Marty’s issue of making
sure direct concatenation can be both simple from a user perspective and
conceptually simple. I’ll post something as soon as I can. =Drummond From: Schleiff, Marty
[mailto:marty.schleiff@boeing.com] Sadly, it seems to me that every time we
raise issues with a recent proposal, we get a new proposal that is even more
complex than the prior one. I agree with Markus that Cross-Resolution is too
complicated. If we're going to change anything at all,
the focus should be on simplification. I'm not opposed to direct concatenation
if we can figure out how to make it simple (including simple from the user
perspective AND conceptually simple). I believe that every subsequent
enhancement to the original direct concatenation proposal introduces new
complexities, and even breaks concepts that were clear in earlier versions. OpenId is looking better all the time.
It's nowhere near as rich, but it's nowhere near as complex. Marty.Schleiff@boeing.com; CISSP
From: Markus
Sabadello [mailto:markus.sabadello@gmail.com] Hey, greetings, On 5/10/07, Drummond
Reed <
drummond.reed@cordance.net > wrote: Bill, |
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]