[gram-user] date/time values in GRAM Audit V2

Tom Scavo trscavo at gmail.com
Tue Apr 21 08:10:51 CDT 2009


On Tue, Apr 21, 2009 at 7:32 AM, JP Navarro <navarro at mcs.anl.gov> wrote:
>
> This doesn't address you storage format issue, but in the context of
> TeraGrid
> information services where we aggregate distributed data with timestamps
> we've
> adopted a specific ISO 8601 format and record all timestamps in UTC
> "coordinated
> universal time" regardless of the local timezone.

That's the approach that SAML, and XML security in general, takes as well.

> All our timestamps look
> like:
>   2009-04-21T12:06:35Z

Exactly.  Do you allow/support higher precision, that is, fractional
seconds?  As you know, time is often recorded as milliseconds after
the epoch.

> We can centrally collect and compare all our timestamps without any
> conversions,
> and only need to convert when someone needs to view timestamps in their
> local
> timezone (we haven't run into a situation yet where that's a requirement).

Thanks for the data point, JP.

Tom

> On Apr 20, 2009, at 9:20 PM, Tom Scavo wrote:
>
>> On Sat, Apr 18, 2009 at 6:24 PM, Tom Scavo <trscavo at gmail.com> wrote:
>>>
>>> I've made significant mods to the "date/time" section in the GRAM Audit
>>> V2 doc:
>>>
>>> http://dev.globus.org/wiki/GRAM_Audit_V2#Date_Time_field_datatype
>>>
>>> After researching this issue, the solution to this problem is even
>>> less clear than it was before.  In fact, I'm leaning towards storing
>>> date/time values as VARCHAR across the board.  At least it's
>>> consistent and permits timezone values.  No other data types have
>>> these properties.
>>
>> Here's another reason to use VARCHAR for datetime columns:
>>
>> https://bugzilla.mcs.anl.gov/globus/show_bug.cgi?id=6566#c34
>>
>> I definitely think VARCHAR is better than TIMEZONE.
>>
>> Tom
>
>


More information about the gram-user mailing list