W3C home > Mailing lists > Public > w3c-dist-auth@w3.org > October to December 2005

Re: XML InfoSet and property value preservation

From: Lisa Dusseault <lisa@osafoundation.org>
Date: Fri, 11 Nov 2005 14:53:57 -0800
Message-Id: <885204fb4832103eb89b400c575a5a5e@osafoundation.org>
Cc: WebDav WG <w3c-dist-auth@w3.org>
To: Julian Reschke <julian.reschke@gmx.de>

In WebDAV implementations today, the only InfoSet piece that we know 
isn't always preserved on properties is the namespace prefix (and thus 
the namespace declaration attributes might also be rewritten).   I'd 
assume we want to encourage new implementations to have at least as 
much preservation/fidelity as existing implementations, and to make 
implementations tend to be consistent with each other.  So why would we 
allow servers to strip out stuff like comments?  If servers reliably 
preserve those today, I don't see why we'd explicitly allow those to 
stop being preserved.

Is there some reason why we would choose to be lax on this in the 
specification?

Lisa

On Nov 6, 2005, at 12:04 PM, Julian Reschke wrote:

> Lisa Dusseault wrote:
>> ...
>>> You're confusing the terminology (that we can re-use) with the 
>>> question what parts of the Infoset WebDAV wants to make 
>>> reound-trippable. Those do not need to be the same. The XML Infoset 
>>> spec just helps in talking about these things.
>> Fair enough, I think the terminology does make things clear and we 
>> can define our own list about what's round-tripped.   On those terms 
>> I'm perfectly in favour of using InfoSet terminology. So does this 
>> boil down to saying something like:
>>   "All Information Items as defined in XML InfoSet [ref] MUST be 
>> round-tripped as part of the property value, with the exception of 
>> the following
>>    - On an element information item: the prefix and namespace 
>> attributes
>>    - On an attribute information item: the prefix
>>    - On a namespace information item: the prefix
>
> Why would we need any Namespace Information Items anyway? I thought 
> namespace declarations weren't considered important, as long as 
> namespace names are preserved for each item?
>
>>   Anything that is not part of the Information Set does not need to 
>> be preserved by the server."
>> ...
>
> So are comments, processing instructions, unparsed entities and so on 
> in? Probably not.
>
> Looking at the large set of information items, it's probably simpler 
> if we just list the items we want to be round-tripped, such as:
>
> 1) On the property element itself: [namespace name], [local name], 
> [children] of type element or character, plus [attributes] named 
> "xml:lang" present on the element itself or it's closest ancestor
>
> 2) On all children of the property element: [namespace name], [local 
> name], [attributes] and [children] of type element or character.
>
> Regarding the issue that started the whole discussion: we IMHO should 
> encourage servers to preserve the [prefix} on all but the property 
> element itself, and warn clients about information loss for those 
> servers that don't.
>
> Best regards, Julian
>
Received on Tuesday, 15 November 2005 01:12:08 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 2 June 2009 18:44:11 GMT