Re: Fixed Section 1.1 language

At 05:52 PM 2002-07-29, Tim Bray wrote:

>noah_mendelsohn@us.ibm.com wrote:
>
>><original>
>>The introduction of new URI schemes SHOULD be avoided.
>></original>
>><proposed>
>>The unnecessary introduction of new URI schemes SHOULD be avoided.
>></proposed>
>>- or - <proposed>
>>The introduction of new URI schemes SHOULD be minimized.
>></proposed>
>>Presumably there are, from time to time, good reasons for introducing new schemes.  Introduction of such schemes SHOULD NOT be avoided IMO.  The original suggests it's always a bad idea.
>
>Yes, but isn't this exactly the semantic of SHOULD - do it unless there's a reason not to?

No, that's not a safe definition of SHOULD.

The problem is not in the interpretation of SHOULD, anyway, but in the 
clause to which it is applied.  'Avoid' new URI schemes is vague to a 
degree detrimental to the purpose.  _You_ may accept those words as matching 
your intended sense, but you can't count on other readers recovering a 
sense compatible with yours from those words.  What it says is true enough
but not precise enough to be helpful.

We should try to articulate the tests that should be satisfied to consider 
a new URI scheme justified.

These have to do with innovations in expressive power, consistency of 
interpretation, and efficiency in operation.  But I don't know where to set 
the threshold surface in that triple-space.

We have a negative example: just to create a namespace.  

Do we have a positive example in the case of binding MIME-compatible type 
indications, complete with parameters, to URI syntax?  Perhaps.

Where does one draw the line in between?

Al


>><original>
>>It is often necessary to compare URIs for equivalence to determine whether they identify the same resource. URI schemes vary in their definitions of equivalence. 
>>For example, URNs
>></original>
>>I understand and agree with what's intended by this statement.  On the other hand, it seems to unintentionally open the broader question of taking two arbitrary URI's and determining using some general means whether they refer to the same resource.  For example, does (http://www.w3.org/TR/REC-xml) refer to the same resource as (http://www.w3.org/TR/2000/REC-xml-20001006).  I believe that, at the moment, it does. 
>><proposed>
>>It is not in general possible to determine by inspection whether two different URI's refer to the same resource.  Particular URI schemes MAY, however, mandate equivalence for particular sets of URIs using that scheme. 
>>For example, URNs ...
>></proposed>
>
>Yep, good catch.  Will fix up & republish. -Tim

Received on Thursday, 1 August 2002 20:52:30 UTC