Re: XML Schema part 2 -Datatypes

Jon:
Thanks for your comments.  I'm delighted that you are using DCD!

We are moving towards simplifying the allowable lexical formats
for datatypes.  Stay tuned!  I will look at the 2nd edition of ISO 8601.

As for i2, i4, r4, etc. you can easily define them as user-defined
datatypes using the max/min facets.

Regards, Ashok


                                                                   
 (Embedded                                                         
 image moved to "Jon Dart" <jdart@tibco.com>                       
 file:          05/26/99 04:53 PM                                  
 pic01483.pcx)                                                     
                                                                   




To:   www-xml-schema-comments@w3.org
cc:    (bcc: Ashok Malhotra/Watson/IBM)
Subject:  XML Schema part 2 -Datatypes




We are using a similar type system (derived primarily from the DCD technical
note)
in a number of our products. We use XML metadata for a somewhat different
purpose
than XML Schema is designed for - we are primarily interested in supporting
object-oriented modeling of business objects. These objects may be realized in
various forms: e.g. as rows in a database table, as messages travelling through
a network, or possibly as XML. Our focus is therefore not so much on specifying
and validating the string representation of types in XML - we are more
interested
in having a robust, general type system that can be applied in various contexts.

Still, we have hit many issues that XML Schema addresses, so I'm sending the
following comments.

1. The text of ISO 8601 is quite imprecise in many areas. There is a draft
2nd edition of ISO 8601 available from
http://www.cl.cam.ac.uk/~mgk25/8601v04.pdf
which tighens up and clarifies much of the language.

2. As you probably know, ISO 8601 has a very wide range of allowable formats
with
many permissible variations. I would recommend specifying a subset of ISO 8601
for use with XML Schema. We have made the following simplifications, for
example:

   a. Dates and times are in "extended format" only, e.g. 1999-05-27, not
19990527;
       23:03:27, not 230327.
   c. Commas are not permitted as "decimal point" indicators.
   d. None of the "truncated representations" are permissible (e.g. "-05-27" for
a date).
   e. Our "interval" data type only allows the "period of time" representation,
for
       example "P2DT10H20M6.6S".

These may be too restrictive; but I still think ISO 8601 (1988) is too imprecise
and
permissive to be a good standard by itself.

3. We use the types "i1", "i2", 'i4", "i8", "ui1", "ui2", "ui4", "ui8" for
signed and unsigned
integers of sizes 1, 2, 4, and 8 bytes. Also "r4" and "r8" for 4- and 8-byte
reals.
I think these were in DCD. It appears you have a more general integer type
system, but it would be nice to have these specific types predefined.

--Jon

-------------------------------------------------------------------
Jon Dart                               jdart@tibco.com
TIBCO Software Inc.              650-846-5099