Actually I think what we have right now is
fine. It’s slightly more restrictive than generic URI syntax, but that’s
ok. I’ll write up the differences and send it to the list tomorrow.
Dave
From: Drummond Reed
[mailto:drummond.reed@cordance.net]
Sent: Monday, February 14, 2005
3:24 PM
To: 'Wachob, Gabe'; Dave McAlpin;
xri@lists.oasis-open.org
Subject: RE: [xri] Relative XRIs
Gabe,
Actually, "//@foo" is exactly
the type of thing I thought Dave was asking about, although he chose to
illustrate his question using an IRI authority rather than an XRI authority. In
other words, he was referring to "//foo.com" instead of
"//@foo", but I thought they both fell into the same count.
But I also
didn't think that any XRI that starts with "//" could be relative. I
thought that character sequence at the start of an XRI always indicated an
absolute XRI. Since "@foo" is already an absolute XRI, my observation
was only that "//@foo" should also be considered an equivalent
absolute XRI.
How can we wrestle this one to the ground
quickly?
=Drummond
From: Wachob, Gabe
[mailto:gwachob@visa.com]
Sent: Monday, February 14, 2005 3:11 PM
To: Drummond Reed; Dave McAlpin;
xri@lists.oasis-open.org
Subject: RE: [xri] Relative XRIs
Drummond -
You aren't suggesting that
//@foo is a legal XRI, are you? Thats really odd. Why would this ever appear?
It leads to a complicated rule: if // is at the beginning, then its relative,
unless the first char after the // is a GCS char.
I'd rather just have
the same rule as URI: if the URI starts with a scheme, then its either a
relative or absolute, depending on the presence of // after the scheme. if the
URI starts without a scheme, then its relative. (Where scheme is scheme chars
followed by a :)
I think Dave at one point
suggested that the first path segment shouldn't be allowed to be empty, and
thats why // could never be a relative XRI. I don't know why Dave had made this
suggestion and I don't see now why we should have this restriction (and thus
why we can't have // begin a relative XRI reference).
-Gabe
From: Drummond
Reed [mailto:drummond.reed@cordance.net]
Sent: Monday, February 14, 2005 2:37 PM
To: 'Dave McAlpin';
xri@lists.oasis-open.org
Subject: RE: [xri] Relative XRIs
I noticed the same thing in my review but
didn't flag it. The one place we diverge from IRI and URI is in the rules
around the opening "//".
I would be in favor of realigning the XRI
2.0 rules so that for an absolute XRI:
1) "xri:" (the scheme name) is
optional,
AND
2) IF an XRI starts with a GCS character,
then "//" is also optional (i.e., the scheme name plus the
"//" can be left off). If the XRI does not start with a GCS
character, then "//" is not optional.
But by these rules the "//" can
always be included.
Mike, is this cool with you?
=Drummond
From: Dave McAlpin
[mailto:Dave.McAlpin@epok.net]
Sent: Monday, February 14, 2005 2:01
PM
To: xri@lists.oasis-open.org
Subject: [xri] Relative XRIs
There’s probably a good reason for this, but why is
“//www.epok.net” a valid relative reference as a URI and IRI but
not as an XRI?