W3C home > Mailing lists > Public > public-qa-dev@w3.org > March 2005

MarkupValidator/M12N: observation/observators

From: olivier Thereaux <ot@w3.org>
Date: Wed, 9 Mar 2005 02:10:32 +0900
Message-Id: <a3b08c3c1188cf5a64dddcdd43c01c92@w3.org>
To: QA Dev <public-qa-dev@w3.org>
A couple of thoughts and suggestions on 
http://esw.w3.org/topic/MarkupValidator/M12N

I have a few minor other changes for the doc (please, tell me 
"misssplled" is on purpose :) but will do these directly in the wiki, 
only including ideas that are probably worth discussing a little here.

The Observation/observer model is very well developed already... The 
main point of my focus for suggestions was what is called "descriptor" 
for an observation (for which I would probably use some term like 
"context").

On "Unfortunately not all input has a network location", I think we 
really need to come up with a way to have an identifier for each 
resource. data: URL scheme is out of question for the reasons described 
in the doc, but I think we could use a "private" URI space with URIs 
that do not dereference to anything, e.g 
http://vwo/id/upload/verylongidentifier (md5?)

The text has a section on how descriptors should be able to identify a 
single character in the source. I would expand this a little further to 
require that a descriptor/context should identify a given range in the 
source (and then a single character would be a specific type of range). 
That's probably not compatible with the current observation model of 
e.g OpenSP.

Speaking of OpenSP, the "Global responsibilities" section asks the 
question of how observations should be of a certain type: "error", 
"violation of foo" (quoting: "This requires the notion of an "error". 
OpenSP does not really provide this notion, even though there is 
currently a 1:1 relationship between OpenSP observations and errors the 
Validator would consider here").

First thought on this matter was that perhaps OpenSP in itself was not 
a "full" observer and should be wrapped in something else to be one, 
but the next section on identifiers and the description of observations 
made me think that maybe this was a non-issue, if each observation id 
could be described as being such and such type of error/violation.

All for now, meeting time...
-- 
olivier

Received on Tuesday, 8 March 2005 17:10:35 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 19 August 2010 18:12:45 GMT