lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Walter Underwood <wunderw...@netflix.com>
Subject Re: Opensearch XSLT
Date Tue, 27 Nov 2007 03:42:29 GMT
"FUD" is pretty strong language. I'll provide some context for
my opinions.

The year before OpenSearch came out, I'd designed and implemented
a SOAP distributed search protocol to go across Ultraseek and
Verity K2, so I was pretty familiar with heterogeneous search
protocols, especially those that had to work for all the extra
features that people depend on.

You can see an example here:

  http://search.ultraseek.com/webservices/

I do *not* recommend using SOAP for protocols. That was a horrific
experience.

Every one of the things I mentioned is something that weakens the
format and would have been fixed with review. OpenSearch was released
as a 1.0, with no prior review. 1.1 remains in Draft 3. It doesn't
look like it has moved forward in two years. Either you implement
1.0 or pretend 1.1 Draft 3 is a finished spec:

  http://www.opensearch.org/Specifications/OpenSearch/Changelog

I tried to look this stuff up in the current spec, but it didn't display
in FireFox 3.0b1, so I checked some of it against the 1.0 spec.

The template parameter you point to appears to be about language
filtering and implicitly about linguistic processing. It would be
much better to separate query processing and results filtering as
the STARTS project did and as we did in the original Ultraseek
distributed search and the SOAP web service.

I really recommend looking at STARTS if you are interested in any
kind of distributed search. It is fundamental work, and the workshop
was a group of active implementers. Heck, Doug Cutting was there.

  http://infolab.stanford.edu/~gravano/workshop_participants.html

wunder

On 11/26/07 6:28 PM, "Ed Summers" <ehs@pobox.com> wrote:

> On Nov 26, 2007 5:35 PM, Walter Underwood <wunderwood@netflix.com> wrote:
>> GData is really pretty useful. OpenSearch was just sloppy. Some element
>> names were capitalized, some weren't. A bunch of stuff specific to A9's
>> UI was mixed in. They insisted on using RSS in addition to Atom for a new
>> application. They supported many encoding types, something that belongs
>> in the presentation layer, not the protocol. [Ever tested a product that
>> supported 150 encodings? I have.] Then they left out any way to specify
>> the natural language used to process the query. They used two different
>> formats for the parameters to a POST and a GET even though HTML has used
>> a single format forever (the <FORM> element).
> 
> Apart from supporting RSS and Atom this doesn't sound much like the
> most recent version of the spec I've looked at [1]. But of course you
> are entitled to your opinion about the "sloppiness". I totally agree
> about the merits of putting standard documents like OpenSearch through
> a process like IETF.  I just think putting down efforts like
> OpenSearch with verbage bordering on FUD doesn't do anyone much good.
> 
> //Ed
> 
> [1] 
> http://www.opensearch.org/Specifications/OpenSearch/1.1#The_.22language.22_par
> ameter


Mime
View raw message