Fwd: Re: Feedback on comparator draft

 >Date: Mon, 06 Dec 2004 17:30:15 -0800
 >From: Chris Newman <Chris.Newman@Sun.COM>
 >Subject: Re: Feedback on comparator draft
 >To: Lisa Dusseault <lisa@osafoundation.org>,Martin Duerst <duerst@w3.org>

 >Lisa Dusseault wrote on 12/6/04 17:11 -0800:
 >> At last IETF I promised to review the comparator draft.  I've now read it and
 >> have three comments or questions, although be warned that I'm not the best
 >> reviewer (I find many more issues when I start using something, than when I
 >> just read it through)
 >>
 >> Section 3.2: I don't understand the requirement for clients supporting
 >> disconnected operation SHOULD NOT use wildcards.  I can imagine a client
 >> synchronizing data with a WebDAV server -- the client would like to download
 >> the contents of a collection in a certain order and cache the results locally
 >> (potentially for offline use).  Why shouldn't this client use wildcard
 >> matching while selecting the collation?
 >
 >For disconnected operation, the client needs to be able to replicate the exact
 >same ordering that the server uses in order to provide a consistent UI.  With a
 >wildcard, the orderings could be different (if, for example, the client had a
 >newer collation than the server in addition to the server-compatible one) and
 >that would not be detected and confuse the user.
 >
 >> Overall, I'm surprised not to see some discussion about providing the
 >> language to be used during collation.  A list of words is sorted differently
 >> in a Spanish dictionary than an English dictionary.  Did I miss something?
 >> has this already been discussed?  They can even be obsolete usages of
 >> collations since these examples would be firmly non-normative.
 >
 >A collation can be designed for a specific language (en;ascii-casemap is one
 >such example).  And the basic comparator can be "tailored" or "customized" for
 >a specific language (creating a new comparator).  Both cases are mentioned in
 >the draft, I believe.
 >
 >> I generally prefer more examples in specifications but I can see how hard
 >> this would be for this spec.  Unless it's possible to briefly illustrate the
 >> use of collations by drawing an example from sieve and/or ACAP...
 >>
 >> As an approach to the problem I worry that it might be overkill but I have a
 >> lot of confidence that you two would know that better than I do.   E.g. if it
 >> were possible to have only one recommended collation then we wouldn't need a
 >> registry or a way to negotiate collations, but I'm sure that approach was
 >> tried first.
 >
 >Your observation about the English and Spanish dictionaries is by itself
 >sufficient justification for a registry.  There is no one-size-fits-all
 >collation and we know it by observation.
 >
 >                - Chris

Received on Wednesday, 18 May 2005 08:07:06 UTC