Gabe,
I thought about this over the weekend, and
I realized that regardless of what we decided, we should recommend that coders maintain
backwards-compatability with XRI 2.0 syntax by accepting XRIs that include BOTH
“xri:” and “xri://”. In fact I think that’s a
recommended practice going forward because regardless of what the spec says, some
users may type “xri:” before their XRI (thinking that’s the
right thing to do) and some may type “xri://” (thinking that’s
the right thing to do) and most will just start at the GCS character (which is
the easiest thing to do) and none of them should have the wrong thing happen.
However since I believe it is important
that XRI architecture maintain the proper ABNF foundation, I think that
formally the best thing is to simplify to “xri:” in XRI 2.1.
So my vote is +1.
=Drummond
From: Gabe Wachob
[mailto:gabe.wachob@amsoft.net]
Sent: Monday, April 09, 2007 10:06
AM
To: 'Barnhill, William'; 'Victor
Grey'; xri@lists.oasis-open.org
Subject: RE: [xri] Moving from
"xri://" to "xri:"
I wish we had done this a long time ago,
to be honest.
HOWEVER, we didn’t, and now we have
the entire OpenID world using XML documents with the namespace
“xri://…”
So we’d be asking the entire world
(whoever that is) to change their XRDs and their implementations to recognize
and produce XML documents in a new namespace. That worries me, especially given
the sentiments some already have toward XRI in OpenID.
So all things being equal, my vote is
actually, because it really hasn’t been that big a problem for the last n
years, so I’m not sure we should pay the price now to “fix”
it. If someone were to convince me the price of the change were actually not
that big a deal (ie someone from the OpenID world said “hey, no problem,
change the XRDs, not a problem”), then I could be persuaded to change to
a +1:
-1 is my vote.
-Gabe
From: Barnhill,
William [mailto:barnhill_william@bah.com]
Sent: Monday, April 09, 2007 8:34
AM
To: Victor Grey;
xri@lists.oasis-open.org
Subject: RE: [xri] Moving from
"xri://" to "xri:"
+1
as well from me.
From: Victor
Grey [mailto:victor@idcommons.org]
Sent: Mon 4/9/2007 10:04 AM
To: xri@lists.oasis-open.org
Subject: Re: [xri] Moving from
"xri://" to "xri:"
+1
=vg
Drummond Reed wrote:
> XRI TC Members and Observers:
>
> In discussion among the XRI Syntax 2.1 editors late last week, the
> question arose as to whether we needed to keep the “//”
after
> “xri:” so that an XRI would be considered to have an RFC
3986/3987-
> compliant authority component. That was the motivation that
“//”
> was added in XRI 2.0 syntax -- XRI 1.0 syntax used only
“xri:”
> followed by the XRI authority component.
>
> We realized that many of the reasons we desired compatability with
> RFC 3986/3987-compliant authority syntax, such as the ability to
> process IP addresses or DNS names in their URI/IRI syntax form,
> have changed in XRI 2.1. For example, the XRI $ Dictionary entries
> $ip and $dns will now be used to express IP addresses and DNS
> names. Also, XRI-normal form will now explicitly not use the scheme
> name, which is only added in IRI-normal form (see http://wiki.oasis-
> open.org/xri/XriCd02/FormsAndTransformations).
>
> We also realized that having the XRI scheme use the URI/IRI “path-
> rootless” rule instead of the previous “authority
path-abemtpy”
> rule would still give us URI-reference normalization, comparison
> rules, etc. But we would gain the following advantages:
>
> 1) An XRI “authority” will not be mistaken by a URI/IRI parser
for
> a URI/IRI “host” (which it sometimes matched and sometimes did
not).
>
> 2) Since "@" is explicitly allowed as a character in a
URI/IRI
> path, we don't have to worry about escaping it when it is not the
> first character in the authority (as we would otherwise have had to
> do to conform with the URI/IRI “host” subrule of the
“authority”
> rule).
>
> 3) We no longer need to specify/support the URI/IRI “port”
rule,
> which has never really applied to XRI.
>
> 4) We can refer to the first segment in an absolute XRI as the
> "authority segment" because it is technically part of the path.
>
> 5) It makes it even clearer that XRI is designed to support
> multiple types of roots, including p2p roots, and XRI roots are
> rooted on GCS chars instead of "//".
>
> We have thus updated the proposed XRI 2.1 ABNF listed at:
>
> http://wiki.oasis-open.org/xri/XriCd02/XriAbnf2dot1
>
> Note that all proposed revisions are logged on this page so you can
> see the entire progression of the editor’s thinking from the
start
> of the ABNF revision process. Please send any questions or feedback
> regarding this proposed change – or any other change – to the
list,
> as we would like to produce Working Draft 01 of XRI Syntax 2.1
> prior to our f2f meeting week after next.
>
> Thank you,
>
> =Drummond
>
>