Re: [xacml] Problems with XACML and time

Steve,

Some of my colleagues agreed on your
interpretation.

Satoshi Hada

IBM Tokyo Research Laboratory 

mailto:

Steve Hanna <>

2003/07/23 23:04
        

        To:
       Satoshi Hada/Japan/IBM@IBMJP

        cc:
       

        Subject:
       Re: [xacml] Problems with XACML and
time

       

Satoshi Hada wrote:

> Here is my understanding:

> 1. The first part of your problem makes sense to me.

For those who are trying to follow along, I think
that

this was the part where XML Schema says that a time with

no time zone cannot be compared to a time with a time zone.

> 2. I don't think the second part is a problem.

>

> >> However, according to the XML-Schema specification, you have
to

> >> normalize the three time values. They are normalized to

> >> 19:00 <= 00:00 <= 03:00. It seems to me that you say
that the

> >> result is false because 19:00 <= 00:00 is false.

>

> I don't think this interpretation is right.

> The XML schema specification says that:

>   The order relation on time values is the Order relation on

>   dateTime (section 3.2.7.3) using an arbitrary date.

>

> So I think that we can use an implicit date information for

> time comparison.

>

> [explanation removed]

The text in XML Schema is unclear. You have an interesting

interpretation, which does solve my problem. However, it has

the strange side effect that 09:00(+14:00) is not equal to

09:00(-10:00). Normally, I would expect them to be equal,

especially since XML Schema says "time represents an instant

of time that recurs every day".

Here are the relevant calculations:

> 1. I want to check whether

> 09:00(+14:00) == 09:00(-10:00)

>

> 2. Set an arbitrary date (e.g., 2002-02-16).

> So I need to check whether

> 2002-02-16T09:00(+14:00) == 2002-02-16T09:00(-10:00)

>

> 3. Normalize the time values.

> So I need to check whether

> 2002-02-15T19:00Z == 2002-02-16T19:00Z

>

> 4. The result is false.

Does this seem odd to you? It does to me. I think
that

any order relation on time values *must* be consistent

with the definition of this data type ("an instant of

time that recurs every day"). That's fundamental.

Therefore, I conclude that when comparing times with
a

time zone there can only be 24 hours of distinct values.

In order to have a total ordering over those values, one

of the values must be chosen as the first. My understanding

of the comparison algorithm is that you normalize to GMT

first, then set the arbitrary date, then do the comparison.

This ensures that there are only 24 hours of values. And

it also means that midnight GMT is the earliest possible

time and 23:59:59 GMT is the latest. I'll repeat my

example above with this algorithm for an illustration.

> 1. I want to check whether

> 09:00(+14:00) == 09:00(-10:00)

>

> 3. Normalize the time values.

> So I need to check whether

> 19:00Z == 19:00Z

>

> 2. Set an arbitrary date (e.g., 2002-02-16).

> So I need to check whether

> 2002-02-16T19:00Z == 2002-02-16T19:00Z

>

> 4. The result is true.

Maybe we need to ask someone who is more familiar
with

the time type in XML Schema what is the proper interpretation

of the text in the spec. Or maybe you argument above

is convincing to you. Please let me know.

Thanks,

Steve