Re: [tdwg-tapir] Proposal

5 Jan 2006

      Hi everybody,
If there are no serious objections or other alternatives to that 
package of changes that was recently proposed, I'll take the liberty 
to commit them in the svn repository. If anyone finds any issues, or 
has any other suggestions, or disagrees with anything, please don't 
be shy!
I'll try to commit the changes individually just to keep a clean 
change log. However this will produce several notification messages 
(sorry!).
Regards,
--
Renato
On 16 Dec 2005 at 16:18, Renato De Giovanni wrote:
...
Dear all,
I would like to propose a set of changes to the protocol trying to 
address the main issues that we've recently discussed here. The 
changes are described below and if you want to browse the 
corresponding schema you can find it here:
http://www.cria.org.br/~renato/tapir.xsd
The main issues are related to TapirLite (how to accomodate it 
without making generic clients overly complicated), and are also 
related to Donald's suggestion about inventory templates. All other 
changes proposed were a mere consequence of browsing the schema and 
finding parts that I thought could be improved.
Please, take a look when you have some time and let me know if you 
have any comments, suggestions, objections, etc. Personally I'm happy 
with the changes, so if you all agree, and if we don't find other 
issues, I'll suggest to "feature freeze" it as soon as possible so 
that people can (re)start implementation based on it.
The feature freeze was originally planned for yesterday (December 
15th), but since the current version still has open issues and this 
proposal should obviously be reviewed and discussed, I'll suggest 
that we postpone it for at least January 5th. By the way, I'll be on 
vacation during the next two weeks, so only accessing e-mail 
sporadically.
Best Regards,
Renato
------- TAPIR / proposed changes -------

Change filter capabilities so that implementations either accept

all filter functionality or none. Specific changes: rename element 
/response/capabilities/operators to "filter" and include an optional 
subelement called "encoding". Rename "boolean" to "booleanOperators", 
and make all logical and comparative operators mandatory. This way 
parsers can always rely on the existence of a <filter> element and 
then pass it to a specific handler (same technique suggested by Steve 
in other parts of the protocol). Such handler can just detect if 
there's an element "encoding" or not, which means "I understand the 
basic filtering mechanism" or "I don't know how to parse any 
filters".
TapirLite implementations will answer in the capabilities:
...
  <filters/>
...
Full TAPIR implementations will answer:
...
  <filters>
    <encoding>
      <expression>
        <concept/>
        <literal/>
        <parameter/>
        <variable/>
        <arithmetic/>
      </expression>
      <booleanOperators>
        <logical>
          <not/>
          <and/>
          <or/>
        </logical>
        <comparative>
          <equals/>
          <greaterThan/>
          <greaterThanOrEquals/>
          <lessThan/>
          <lessThanOrEquals/>
          <in/>
          <isNull/>
          <like/>
        </comparative>
      </booleanOperators> 
    </encoding>
  </filters>
...

Change multiCOPType (basis of the <in> operator) so that <values>

can contain a sequence of "expression" elements instead of only 
literals. This will allow parameters with multiple values (or 
multiple parameters) to be used by the TCS API.

Re-include an operation completely dedicated to the usage of

queryTemples (I called it "view" but I'm happy with any other better 
names if people feel uncomfortable with it). This operation would 
only be accessible through GET whithout using XML requests. And it 
would not accept the "filter" and "partial" parameters. This seems to 
be the cleanest way to accomodate TapirLite since the _only_ change 
in the schema is the new optional element 
/response/capabilities/operations/view indicating the query templates 
supported. Obs: I tried other ways, but the results were not so nice 
and the changes were more drastic.

Remove the possibility to use queryTemplateURI in XML search

requests. If someone wants to use a queryTemplateURI in a search 
request, it's so much simpler to use regular GET parameters in the 
URL that it's looks reasonable to simplify the protocol in these 
situations.

Move /response/capabilities/operations/search/queryTemplates to

/response/capabilities/operations/view/queryTemplates.

On the other hand, include a new <predefinedOutputModels> element

in /response/capabilities/operations/search (issue raised in a recent 
discussion with Rob in this mailing list). DiGIR2 will make use of 
it.

Reduce "queryTemplateGroup" to contain only <label> and

<documentation>. Move the other elements to a new 
"searchTemplateGroup" which also references the "queryTemplateGroup". 
Create a new "inventoryTemplateGroup" in the same way, and make the 
inventory request element reference it (as it happens in the search 
request definition). Remove global element <queryTemplate>, including 
<searchTemplate> and <inventoryTemplate> to make the schema validate 
external template definitions. This way query templates used in view 
operations can point to predefined search or inventory operations.
Other changes:

Make /response/capabilities/concepts/schema a mandatory element so

that any implementation shows which concepts are present.

Remove controlled vocabulary of element

/response/metadata/relatedEntities/entity/contact/role and make it 
unbounded: controlled vocabularies do help with standardization, but 
they also reduce flexibility. In different contexts people could 
easily envision different uses for that attribute, so I would rather 
enforce vocabulary control in other levels (such as software 
configuration or network configuration).

Remove "datum" attribute from geoLocationType: the corresponding

element has been included so that "related entities" (part of 
metadata responses) could indicate their location. People can easily 
always put lat/long in WGS84 in this case, it's just a matter of 
adding a note about that in the documentation.

Rename primaryConceptNameServer and secondaryConceptNameServer to

primary and secondary, respectively (the enclosing element already 
provides the context). Make <primary> optional.

Rename global element <environment> to <variable>, and move

/response/capabilities/search/variables to 
/response/capabilities/variables since they can be used in both 
mappings and filters. Remove redefinition of all variables in the 
capabilities response by referencing the existing variable element 
defintion. Remove empty 
/response/capabilities/search/variables/environment/metadata element.