comments on Expires

[Will Norris <will@willnorris.com>]

Joe Gregorio and Martin Atkins are asking on the xri-comments list for more clarification on the Expires element.  Particularly, two concerns were brought up:

 - what should an XRD consumer do when a document is expired?  obviously it should try and fetch a new copy of the document, but how strong of language do we want to use here?  "Consumers MUST NOT process an expired XRD" ?

 - what happens when the <Expires> date has passed, but the HTTP expires date has not yet passed?  Depending on your HTTP stack, when you attempt to fetch a fresh copy of the document, you may very likely end up with the same document.  Perhaps a recommendation for XRD providers to prevent this from happening?


It's certainly worth noting that XRI Resolution 2.0 explicitly addressed both of these cases.  While I do prefer our general approach trying not to over specify the relationship between <Expires> and any transport caching semantics, I do think this is worth addressing.

Thoughts?

-will


On Jan 11, 2010, at 10:47 PM, Drummond Reed wrote:

> On Mon, Jan 11, 2010 at 11:17 AM, Joe Gregorio <joe@bitworking.org> wrote:
> 
>> On Mon, Jan 11, 2010 at 2:02 PM, Will Norris <will@willnorris.com> wrote:
>>> 
>>> On Jan 11, 2010, at 10:44 AM, Joe Gregorio wrote:
>>> 
>>>> I've reviewed the XRD 1.0 spec and have the following issues:
>>>> 
>>>> 1. Expires
>> http://www.oasis-open.org/committees/download.php/35678/xrd-1.0-wd11.html.#element.expires
>>>> 
>>>> "This dateTime value indicates the time instant after which the
>>>> document is no longer valid."
>>>> 
>>>> There is no explanation of what 'valid' means. I presume it is not
>>>> used in the context of XML, so
>>>> it is not referring to XML validity. Should none of the information
>>>> in the XRD document be
>>>> used after the expiration time? Or only the signed information?
>>>> 
>>>> What is the relationship between XML validity and this expires
>>>> validity? That is, if I receive an
>>>> XRD document that does not conform to the schema, obviously that
>>>> means it is XML invalid,
>>>> but does that XML invalidity mean that the document is also invalid
>>>> in the 'expires' sense?
>>> 
>>> I'm actually working on this now.  I noticed we had a slight
>> inconsistency in the spec, as to whether the document was still valid (in
>> the expiration sense) at the exact time instant specified by <Expires>.  In
>> clarifying that, I've also changed the wording:
>>> 
>>>   The <Expires> element contains a dateTime value which indicates the
>> time instant at which the document has expired.
>> 
>> Avoiding the word 'valid' does avoid the ambiguity with XML validity.
>> On the other
>> hand, it does not answer the question of what, in this case, 'expired'
>> means. Should
>> none of the information in the XRD file be used? Should a client try
>> to retrieve a
>> fresher copy?
>> 
> 
> I agree with Joe that the spec should state what the client (XRD consumer)
> either SHOULD or MUST do when the XRD has expired. I believe the simple
> answer is that the client MUST retrieve a new copy.
> 
> =Drummond