RE: [046-lc-edit-relative-URI] proposed patch

I thought I would survey existing RFCs to see if the notion that
the term 'URI', used without any qualification, is used with
a sense that does not include relative references.

This is a 'running code' argument. I just grepped RFCs for the word URI,
 and look at the uses, in reverse order.

Based on this sample, I claim that 'common usage' is that
the term 'URI' use is more consistent with the definition
"does not include relative path" than not.


RFC 3870: no relative paths;
produces an 'absolute URI'.

RFC 3864: no relative path
      e.g.  A postal address, home page URI, telephone and fax
      numbers may also be included.

'home page URI' assumes absolute, a relative reference makes
no sense.

RFC 3863:  no relative path
   All elements and some attributes are associated with a "namespace",
   which is in turn associated with a globally unique URI.  Any
   developer can introduce their own element names, avoiding conflict by
   choosing an appropriate namespace URI.


RFC 3862: no relative path
The BNF is:
   URI          = <defined as absolute-URI by RFC 2396>

RFC 3861: no relative path
  There is no application of relative paths with IM URIs.

RFC 3860: no relative path
  registers IM URIs, no mention of relative paths

RFC 3859: no relative path
  registers IM URIs, without any use for relative paths

RFC 3858: no relative path 
      resource: This attribute contains a URI for the resource being
      watched by that list of watchers.  It is mandatory.


Do I need to go on? The preponderance of 'current practice'
in IETF published documents is that 'URI' by itself without
qualification means 'no relative path'.


Received on Friday, 17 September 2004 22:42:37 UTC