[MKSearch-dev] RESTful query syntax

[Phil Shaw]

On 24 Nov 2004, at 10:42, Bruno Postle wrote:

> There is a mock-up of a RESTful interface for a proposed advanced
> MKDoc search engine here:
> 
>   http://testers.mkdoc.com/ideas/search/

First, I would like to make a controversial suggestion that MKSearch 
does not index any document content, only metadata in meta and link 
elements. Various arguments in favour:

1. No ambiguity about what will be indexed and what is not (or how 
so). If items cannot be found, it will be because they have not been 
catalogued properly, or the query is inappropriate.

2. Such a system therefore (ideally) serves as a metadata review 
tool, which seems to be much needed to me. It would be relatively 
easy to get a metadata snapshot of any document in the database.

3. This gives the system precision in terms of the RDF output that we 
generate. Since the metadata must have been created with explicit 
intent, we can make definite statements about the meaning of the 
results in RDF. This cannot be said of general body or title content.

4. It distinguishes the tool compared with most for a very 
specialised purpose. That is, it is not a general purpose search 
engine, it is a metadata or semantic Web search engine.

5. Reduces storage, speeds performance, simplifies the rules that are 
applied to the content. That means no special treatment of common 
"stop" words that need not be indexed, making phrase searches around 
those, etc.

> This proposes a URI canonicaliser that rewrites & as ; and %3A as =.  

This is not a technical problem, but I would like to be clear about 
doing this for the right reasons. This is me being devil's advocate.

Why do we want the search results page to re-form the original query 
in one field? If someone wanted to adjust a multi-field metadata 
search wouldn't they want a replica of the original search form pre-
filled?

If people want to save or bookmark queries won't they just copy them, 
not reconstruct them by hand? Is there some other purpose you have in 
mind?

I think it could be confusing re-forming the query in a different 
syntax only to rewrite the resulting URL again.

> Presumably a canonicaliser should perform reordering of the query
> string too.

Again, I don't see this as a technical requirement. Is there any 
advantage in usability terms?

I'm a little wary of changing what the user puts into the system in 
unexpected and possibly unwarranted ways.

Best regards,

Phil