Type Detection for RDF/XML files

["Sina K. Heshmati" <si...@khakbaz.com>]

I'm currently using the following pattern to catch the target files for the SKOS plugin:

<map:match pattern="**skos.xml">
...
</map:match>

As the pattern suggests, files whose name ends with 'skos.xml' will be processed as SKOS files. Fine, but how bad is it?

The issue FOR-1005 [1] suggests that the 'sourcetypeaction' mechanism should be used 
to determine the type of a file and set actions accordingly. We see in FOR-1018 [2] 
that we need to extend our existing implementation of the sourcetypeaction mechanism 
to be able to support RDF/XML formats.

One might take an approach similar to bootstrapping, that is:

0. Try to determine the file type by parsing the filename; if failed,
1. Fallback to sourcetypeaction mechanism with a minimum number of tests; if failed,
2. Type left undetermined, that is no actions.

If we decide to parse the filename to determine the type of a file, then we'll need to have conventions set for each format.

What do you think should be done to determine the type of RDF/XML formats?

SinDoc

[1] http://issues.apache.org/jira/browse/FOR-1005
[2] http://issues.apache.org/jira/browse/FOR-1018

Re: Type Detection for RDF/XML files

[Ross Gardler <rg...@apache.org>]

On 10/07/07, Sina K. Heshmati <si...@khakbaz.com> wrote:
> I'm currently using the following pattern to catch the target files for the SKOS plugin:
>
> <map:match pattern="**skos.xml">
> ...
> </map:match>
>
> As the pattern suggests, files whose name ends with 'skos.xml' will be processed as SKOS files. Fine, but how bad is it?

It means that all requests need a fixed URL space. Forrest should not
define a users URL space. However, keeping things moving is important
at this stage.

> The issue FOR-1005 [1] suggests that the 'sourcetypeaction' mechanism should be used
> to determine the type of a file and set actions accordingly. We see in FOR-1018 [2]
> that we need to extend our existing implementation of the sourcetypeaction mechanism
> to be able to support RDF/XML formats.
>
> One might take an approach similar to bootstrapping, that is:
>
> 0. Try to determine the file type by parsing the filename; if failed,
> 1. Fallback to sourcetypeaction mechanism with a minimum number of tests; if failed,
> 2. Type left undetermined, that is no actions.
>
> If we decide to parse the filename to determine the type of a file, then we'll need to have conventions set for each format.
>
> What do you think should be done to determine the type of RDF/XML formats?

In an ideal world we will always avoid using filename. However, there
is a need to progress with your work and a good enough solution is,
well, good enough, until someone has the desire to fix it.

I would say your 0. and 1. steps should be reversed.

Ross