Identifier normalization

[Ray Plante]

Hi gang,

I'm working to get the first version IVOA Working Draft for Identifiers 
finished.  You can see this work in progress at 
http://rai.ncsa.uiuc.edu/~rplante/VO/WD-ID.html.  If you are not familiar 
with the ID proposal as discussed on this list, have a look at the 
"Overview" section in the above draft.  You can also consult an informal 
summary of the proposal via http://www.ivoa.net/forum/registry/0349.htm.

I wanted to solicit some opinions regarding some details of what 
precisely may appear in a resource key.  In particular, some URIs (e.g. a 
URL) allows the following in the path portion:

   o  an optional trailing "/" 
   o  "//" which is equivalent to "/"
   o  "/./" which is equivalent to "/"
   o  "/../"  which means "go up one level"

Before the URI can be resolved, it must be converted to a "canonical" form 
which does not contain any of the above.  This transformation is usually 
referred to as "normalization".

This in principle could apply to resource keys as well.  In this case, the 
transformation would be necessary not only when resolving VO identifiers 
to resource descriptions, but also when comparing two identifiers to see 
if they refer to the same thing.  Thus, to do a full test of whether two 
identifiers refer to the same thing is slightly more work than simple 
string comparisons.

An important question is whether providers think they will need to ".." to 
refer to a resource relative to another.  In light of that, should we:

  *  disallow the above patterns completely,
  *  allow them but treat them literally; i.e. "normalization" is not 
       supported, or
  *  fully support the special meaning of these patterns and support
       normalization during resolution or comparison?

Your comments on this topic or anything you see in the draft thus far are 
welcome.  I hope to have the full draft ready later this week.

thanks,
Ray