GCS chars and star in XRI authority resolution

["Drummond Reed" <drummond.reed@cordance.net>]

Per the question Mike raises: "The inclusion (or not) of the delimiter that
indicates reassignability or persistence in the sub-segment that gets
resolved is something that we'll still need to discuss as we revisit
resolution for 1.1."

I'd like to propose something I think will make both equivalence and XRI
authority resolution simpler in 1.1.

Per the BNF I just posted following Dave's suggestion, a GCS-rooted
authority segment would be:

GCS-authority    =  gcs-char [ "!" / "*" ] [ nz-segment ]

Therefore a GCS char can be followed by either an nz-segment, or !, or *. So
the following are all legal:

@foo
@*foo
@!foo

In XRI 1.0 we treated "@foo" and "@*foo" as equivalent. We said the * was
"assumed" with any GCS char. 

In XRI 1.1 I'd propose that we simplify things in one of two ways:

OPTION 1: BY NOT DEFINING GCS-CHAR AND GCS-CHAR* AS EQUIVALENT

Instead, the rules would be that:

1) By default, the nz-segment following a GCS char is reassignable.
2) In XRI authority resolution, if EITHER ! or * preceed an nz-sub-segment,
they are treated as part of the nz-sub-segment from the standpoint of
resolution, i.e., are part of the value being resolved.

By these rules, @foo, @*foo, and @!foo are all different values. "foo" is a
reassignable sub-segment in both "@foo" and "@*foo" by definition, but
"@foo" and "@*foo" are not equivalent.

OPTION 2: BY NOT ALLOWING * DIRECTLY AFTER A GCS CHAR

In ths option, the BNF would be:

GCS-authority    =  gcs-char [ "!" ] [ nz-segment ]

Again, the same two rules proposed in Option 1 would apply. Only now you can
just have "@foo" and "@!foo", because "@*foo" is illegal.

I believe this is actually the option most consistent with the rule that by
default, the nz-sub-segment following a GCS char is reassignable, because it
means that the * is already inherent in the GCS char, just the way it is
inherent in a slash (reassignable sub-segment being the default after
slash).

Which do folks prefer?

=Drummond 


-----Original Message-----
From: Lindelsee, Mike [mailto:mlindels@visa.com] 
Sent: Friday, August 20, 2004 2:10 PM
To: xri@lists.oasis-open.org
Subject: RE: RE: [xri] Initial proposed XRI 1.1 ABNF and issues analysis

Bill,

I'm not clear which XRI ABNF you are asking your questions with respect to.
I don't believe that the XRI below (xri:@example*:23:45) would be valid
using XRI 1.0 syntax ('*' is only allowed as a GCS character in 1.0).  The
XRI also wouldn't be valid in the original 1.1 ABNF ('*' and ':' can't
follow one another). It would be valid in the various iterations of the ABNF
that Dave, Drummond and I have been discussing on the list -- though the
interpretation of the sub-segments might be slightly different between the
various iterations.

The latest proposal would break the XRI up as follows:

1:  @
2:  *example   (reassignable sub-segment - and showing implicit delimiter)
3:  *:23:45    (reassignable sub-segment)

The inclusion (or not) of the delimiter that indicates reassignability or
persistence in the sub-segment that gets resolved is something that we'll
still need to discuss as we revisit resolution for 1.1.

Mike

> -----Original Message-----
> From: Barnhill William [mailto:barnhill_william@bah.com]
> Sent: Friday, August 20, 2004 11:03 AM
> To: Lindelsee, Mike 
> Cc: xri@lists.oasis-open.org
> Subject: Re: RE: [xri] Initial proposed XRI 1.1 ABNF and 
> issues analysis
> 
> 
> Looks good to me as well, but some questions...
> (1) Is this XRI valid? xri:@example*:23:45
> (2) If valid, would it represent 4 resolution steps:
> 1:   @
> 2:     .example
> 3:        *:23
> 4:           *:45
> With the '. on 1: and the '*' on 4 being implicitly stated.
> 
> (3) If the above XRI is suppose to respresent 4 resolution 
> steps do not 
> the new rules result in only 3 steps? As :23:45 would be 
> considered one 
> segment.
> 
> Thanks,
> 
> Bill Barnhill
>