All,
as the last outstanding item before
closing on the final V1 of WS-Agreement, we need to settle on the use of
strings, QNames and NCNames for descriptive names and IDs that we use in
the schema and spec. Right now, it is a merry mixture, following various
trends in the past.
We have a number of options to address
the restrictions that should apply to unique identfiers and desriptive
names:
- Unique identifiers could be strings
(no restriction), QNames (one can/must use namespaces to structure ids),
or NCNames (no QName). There are pros and cons to all of them but the most
likely candidates seem to be strings or QNames. I don't see implementation
reasons for NCNames right now.
- Descriptive names could be strings
or we may force NCNames. I don't see the point of telling people how they
would like to call their agreement etc. descriptively. If a users thinks
a:b:c tells him or her what is meant who wants to object. So I strongly
propose the first.
Applying the above thoughts to the schema,
we yield:
- a schema where all ids are QNames
and all descriptive names are strings:
- a schema where all ids and names are
strings:
To decide what is an id and what is
a descriptive name I strictly went by the name of the attribute, finishing
with Id or not. This might not be correct. For example, ServiceName coul
dreally be an ID. Then we should call it that way and make it whatever
we decide. If we go with string ids, it doesn't matter.
This point is to be concluded in tomorrow's
phone call and then applied to the text of the spec in the following day
or two.
Heiko
-----
Heiko Ludwig, Dr. rer. pol.
IBM TJ Watson Research Center, PO Box 704, Yorktown, NY, 10598
hludwig@us.ibm.com, tel. +1 914 784 7160, mob. +1 646 675 8469
http://www.research.ibm.com/people/h/hludwig/