Ok, please let me know if you need me to clarify anything in the spec. I'm happy to help where I can. Please also note that I checked in a bunch of tests relating to viewmodes today.

Kind regards,
Marcos

On 7/7/10 6:39 PM, Ricardo Varela wrote:
hallo Marcos (and sorry for the confusion in copying groups)

I think the clarifications below should be fine. We are using the W3C
tests but just wanted to be sure we were interpreting the test cases
in the proper way

Thanks for your help

Saludos!

---
ricardo

On Fri, Jul 2, 2010 at 11:00 AM, Marcos Caceres<[email protected]>  wrote:
Hi Ricardo,

(moving discussion to public-webapps)

On 7/2/10 5:56 AM, Ricardo Varela wrote:

hallo all, hallo Marcos,

We have a small question regarding what we interpret may be an
inconsistency in the behaviours for parsing a config file as commented
in the W3C widget packaging spec [1]

According to the spec (latest and also older versions), the
occurrences of some elements (eg: author or content) have to be zero
or one

I'm sorry, the specification is unclear. It says "expected children (in any
order)", but it certainly is not intended to be a restriction on authors -
that is to say, it would make no sense to punish authors who put in two
author elements by mistake. A conformance checker could then warn if
something out of the expected (such as two author elements) if found in the
document. This is defined in this yet to be published spec:

http://dev.w3.org/2006/waf/widgets-pc-cc/Overview.src.html

However, on the algorithm to process a configuration document quoted
below, it states: "If this is not the first author element
encountered, then the user agent must ignore this element and any
child nodes" It just says ignore and doesn't say to consider it as
error

Isn't this a contradiction in the parsing of the configuration
document? We understand that it should be one of these 2 cases:

a) we allow for more than one instance of author and content and let
the first one take precedence (and therefore the occurrences should be
"zero or more")

No, only one is expected.

b) we allow only one instance of author and content elements (and
therefore the parsing algorithm has got to stop with error on further
occurrences)

Certainly not: the parser is not a conformance checker. The parser should be
able to flexibly handle all garbage input gracefully, as well as be future
compatible (in case we want to allow more than one author or content element
on the future).

Would appreciate some clarification about this, as we want to clarify
what to do for our compliance tests

I hope that clarifies things. If not, I'm happy to discuss further.

Also, are you making your own compliance tests or using the official ones?:

http://dev.w3.org/2006/waf/widgets/test-suite/

Thanks a lot in advance!

Saludos!

[1] http://www.w3.org/TR/widgets/


--
Marcos Caceres
Opera Software





--
Marcos Caceres
Opera Software

Reply via email to