Re: [thredds] More questions about serviceName, urlPath, and access

• To: Nathan Potter <ndp@xxxxxxxxxxx>
• Subject: Re: [thredds] More questions about serviceName, urlPath, and access
• From: Ethan Davis <edavis@xxxxxxxxxxxxxxxx>
• Date: Mon, 07 Dec 2009 13:06:33 -0700

Hi Nathan,

Nathan Potter wrote:
> 
> Ethan,
> 
> Is it allowed for a dataset to have the serviceName and/or the urlPath
> specified as child elements? (I think the answer is yes to serviceName
> and no to urlPath)

That is correct. There is a "serviceName" element; there is no "urlPath"
element.

> Examples:
> 
>    <dataset name="level2_b" ID="nc/level2_b.nc" >
>       <serviceName>s2</serviceName>
>       <urlPath>nc/level2b_.nc</urlPath>
>    </dataset>
> 
> 
>    <dataset name="level2_b" urlPath="nc/level2_b.nc" ID="nc/level2_b.nc" >
>       <serviceName>s2</serviceName>
>    </dataset>
> 
> 
>    <dataset name="level2_b" serviceName="s2"  ID="nc/level2_b.nc" >
>       <urlPath>nc/level2b_.nc</urlPath>
>    </dataset>
> 
> 
> 
> Is it allowed for a dataset to have both a serviceName/urlPath pair as
> attributes AND as child elements?
> If so, which has priority?  (My guess is that none of these are allowed)

The XML Schema does not expressly forbid a dataset having both a
"serviceName" child element and a "serviceName" attribute. However, the
"serviceName" attribute is documented as deprecated in both the schema
and the specification document:

http://www.unidata.ucar.edu/projects/THREDDS/tech/catalog/v1.0.2/InvCatalogSpec.html#dataset

So, we don't write datasets that use the "serviceName" attribute but we
don't choke when reading a dataset with the attribute or both the
attribute and element (our code uses the element if it is avialable).


> Examples:
> 
>    <dataset name="level2_b" serviceName="s2" urlPath="nc/level2_b.nc"
> ID="nc/level2_b.nc" >
>       <serviceName>s3</serviceName>
>    </dataset>
> 
> 
>    <dataset name="level2_b" serviceName="s2" urlPath="nc/level2_b.nc"
> ID="nc/level2_b.nc" >
>       <serviceName>s3</serviceName>
>       <urlPath>data/level2b_.nc</urlPath>
>    </dataset>
> 
>    <dataset name="level2_b" serviceName="s2" urlPath="nc/level2_b.nc"
> ID="nc/level2_b.nc" >
>       <urlPath>data/examples/level2b_.nc</urlPath>
>    </dataset>
> 
> 
> Is a dataset allowed to have a serviceName/urlPathr via inheritance,
> attribute, or children and also have access element children?

Yes.

> If so are there rules about precedence? Is it cumulative?

It is cumulative. A dataset can have multiple access methods. That is, a
dataset can define multiple service/urlPath combinations that each
define a service type and an access URL. Each access method provides
access to the same dataset.

There are two ways for a dataset to define its access methods:

1) The dataset's "urlPath" attribute value is combined with the
dataset's "default" service [*]. If that service is a "Compound"
service, this results in multiple access methods/URLs.

2) Each child "access" element results in a single access method/URL. A
dataset can have any number of "access" elements; the service referenced
by an "access" element can be a "Compound" service. [If the "access"
element does not include a "serviceName" attribute, the dataset's
"default" service [*] is used.]

[*] By "default" service, I mean the service named by the highest
precedence local or inherited "serviceName" value. Precedence:
  1) Local
     a) child "serviceName" element (XPath: "./serviceName")
     b) child "serviceName" element of a child "metadata" element
        (XPath: "./metadata/serviceName")
     c) "serviceName" attribute (XPath: "@serviceName")
  2) Inherited:
     - the closest ancestor "dataset" with an inherited metadata
       that contains a serviceName element (XPath-ish: the first
       item in the set given by
       "ancestor::dataset/metadata[@inherited=true]/serviceName")

Some examples and more details (though, without the precedence
discussion, yet) are given in the "Dataset Access Methods" section of
the catalog spec document:

http://www.unidata.ucar.edu/projects/THREDDS/tech/catalog/v1.0.2/InvCatalogSpec.html#Dataset_Access_Methods

> Examples:
> 
>  <dataset name="level1" ID="degenerate examples">
> 
>    <metadata inherited='true'>
>       <serviceName>s1</serviceName>
>    </metadata>
> 
>    <dataset name="level2_a" urlPath="nc/level2_a.nc" ID="nc/level2_a.nc">
>       <access serviceName="s2" urlPath="nc/level2_a" dataFormat="NetCDF" />
>    </dataset>
> 
>    <dataset name="level2_b" serviceName="s2" urlPath="nc/level2_b.nc"
> ID="nc/level2_b.nc" >
>       <access serviceName="s3" urlPath="nc/level2_a" dataFormat="NetCDF" />
>    </dataset>
> 
>    <dataset name="level2_b" urlPath="nc/level2_b.nc" ID="nc/level2_b.nc" >
>       <serviceName>s2</serviceName>
>       <access serviceName="s3" urlPath="nc/level2_a" dataFormat="NetCDF" />
>    </dataset>
> 
>  </dataset>
> 
> 
> 
> 
> Thanks for all of your help with this!

No problem. Hope I didn't muddy the waters even more.

Ethan

> 
> 
> N
> 
> 
> = = =
> Nathan Potter                        ndp at opendap.org
> OPeNDAP, Inc.                        +1.541.231.3317

• References:
  • [thredds] More questions about serviceName, urlPath, and access
    • From: Nathan Potter