[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [Openxri-users] Proposal to standardize default value for append attribute
Wil, I agree with your first point and
have made a note to add that text to Working Draft 11. Regarding the second one, about what the
default value should be, this is one of those places where I simply feel we
made a mistake in Working Draft 10 and we should correct it as early as we can
or else more and more XRDS developers and implementations are going to be
written assuming a default value of the append attribute that was a poor
choice. In other words, nip the problem as early
in the development stage as possible. This is *exactly* the kind of problem that Working Draft 10 is supposed
to uncover. Thankfully, it’s the only normative issue that’s come
up so far – everything else is simply non-normative changes to text to
better understand/clarify the spec. How do others feel? Is this a problem we
should “nip in the bud”? =Drummond From: William Tan
[mailto:dready@gmail.com] Drummond, On 4/29/06, Drummond
Reed <drummond.reed@cordance.net>
wrote: Wil (and OpenXRI users and XRI TC members both): This OpenXRI thread, which I am cc'ing to the XRI TC because
it is relevant there too, has reminded me of an issue that I too ran across
when writing up XRDS documents. Currently we specify (in section 8.4 of XRI Resolution 2.0
Working Draft 10, http://www.oasis-open.org/committees/download.php/17293/xri-resolution-v2.0-wd-10.pdf
), two different defaults for the "append" attribute. The default value "none" is specified for authority
resolution and the default value of "local" is specified for service
endpoint selection. Even though I helped write that section, it has since struck
me that: a)
It is unusual, potentially confusing (and not
very wise) to specify different default values for an attribute. b)
The more logical and intuitive default behavior
of an absent or null "append" attribute value would be
"none". The reason we specified that "local" should be the
default behavior if the append attribute was absent or null was that we
anticipated that this would be the vast majority of use cases, i.e., passing on
qxri paths to "dumb" web servers. However what I have found in
working on XRDS documents since we completed Working Draft 10 is that this in
fact is rarely the case, i.e., "local" is not chosen any more than
any other option. So my proposal is simple: in Working Draft 11 (which we want
to become the Committee Draft specification), we change section 8.4 to specify
that the default value of the "append" attribute if it is absent or
null is "none". Thus if an XRDS author wants a service endpoint URI
to have the local part of the qxri appended, the author needs to explicitly
specify the value of the append attribute = "local". If anyone disagrees, please post. =Drummond From: openxri-users-admin@lists.sourceforge.net
[mailto:openxri-users-admin@lists.sourceforge.net] On Behalf Of William Tan Hi Mohit, On
4/28/06, Mohit Raj <mohitraj@gmail.com>
wrote: Wil, Here's another tweak I made to the code to get it running.
Though not a big one, just a check more. In the Resolver.resolveSEPToURIList(), append
attribute is only being checked for its presence and the attribute value
is not being checked. Therefore, if the attribute is present but having the
value as append="none"
will do the same thing as append
having any other value. In this am referring to the section 8.4 of WD 10
listing the append attribute values, dealing with the construction of Service
Endpoint URIs. I made the following change: //if (append == null) if (append == null || append
== "none") Somebody pl. validate keeping in mind my circumscribed
knowledge of this. Thanks, Mohit
|
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]