"opensearch.xml" check please

[Andy Mabbett]

I've just placed an "opensearch.xml" [1] file on a website I mange. At
present, it's only referenced from one page:

http://www.westmidlandbirdclub.com/new.htm

and the results are HTML, not RSS/ Atom.

Would someone with IE7 kindly check it out for me, please?

Thank you.

[1] OpenSearch: <http://opensearch.a9.com/spec/1.1/>

--
Andy Mabbett
Say "NO!" to compulsory ID Cards: <http://www.no2id.net/>

Free Our Data: <http://www.freeourdata.org.uk>

[Dan]

Andy Mabbett wrote:

I've just placed an "opensearch.xml" [1] file on a website I mange. At
present, it's only referenced from one page:

http://www.westmidlandbirdclub.com/new.htm

and the results are HTML, not RSS/ Atom.

Would someone with IE7 kindly check it out for me, please?

Which link on that page is the one you're trying to test? And what
does it do in other browsers such as Firefox and Mozilla SeaMonkey?

And why does your self-proclaimed registered charity have a .com
address, implying that it's a commercial entity?

--
Dan

[Andy Dingley]

Andy Mabbett wrote:

I've just placed an "opensearch.xml" [1] file on a website I mange.

Why? This would appear to be a fundamental mis-use of the purpose of
OpenSearch

OpenSearch is an intersting idea - crudely put it's to re-implement the
Z39.50 distributed search approach with a modern XML / RSS based
protocol. It's _not_ intended for publishing static site metadata, as
you appear to be using it.

[Andy Mabbett]

In message <11**********************@h48g2000cwc.googlegroups .com>, Andy
Dingley <di*****@codesmiths.comwrites

>I've just placed an "opensearch.xml" [1] file on a website I mange.

Why?

Because I wanted to.

>This would appear to be a fundamental mis-use of the purpose of
OpenSearch

OpenSearch is an intersting idea - crudely put it's to re-implement the
Z39.50 distributed search approach with a modern XML / RSS based
protocol. It's _not_ intended for publishing static site metadata, as
you appear to be using it.

Well, that's certainly one opinion, but it's not my reading of either:

http://www.last-child.com/add-opense...your-web-site/

or

http://opensearch.a9.com/

--
Andy Mabbett
Say "NO!" to compulsory ID Cards: <http://www.no2id.net/>

Free Our Data: <http://www.freeourdata.org.uk>

[Andy Dingley]

Andy Mabbett wrote:

OpenSearch is an intersting idea - crudely put it's to re-implement the
Z39.50 distributed search approach with a modern XML / RSS based
protocol. It's _not_ intended for publishing static site metadata, as
you appear to be using it.

I was rather hasty in describing your particular OpenSearch example,
because you _are_ using it correctly -- You do have one sub-element in
there that points to a 3rd party search engine, and that's pointed
specifically at your site. You're in a grey area here: _Should_ such
3rd party searches fall under the remit of what OpenSearch describes?,
but thinking about yours more carefully, I now think you're on the
right side of things.

This isn't a _substantial_ use of OpenSearch and it's still up for
debate if this is the best was to do it -- but it's not _wrong_ to do
it this way.

Well, that's certainly one opinion, but it's not my reading of either:

It's still very much my reading of it. OpenSearch is only there to
describe dynamic search engines and it's a mis-use of it to use it to
merely describe static site metadata. Just because the properties it
offers _can_ describe static metadata, that doesn't justify the use of
the protocol if that's all it's being used for.

http://www.last-child.com/add-opense...your-web-site/

"create an XML document that defines your site's search engine
methods"

Note the definition carefully. It's about defining search engines, not
sites. If your site doesn't have a search engine (and you're not
borrowing a 3rd party) then don't use it.

[Andy Mabbett]

In message <11**********************@b28g2000cwb.googlegroups .com>, Andy
Dingley <di*****@codesmiths.comwrites

>Andy Mabbett wrote:

>OpenSearch is an intersting idea - crudely put it's to re-implement the
Z39.50 distributed search approach with a modern XML / RSS based
protocol. It's _not_ intended for publishing static site metadata, as
you appear to be using it.

I was rather hasty in describing your particular OpenSearch example,
because you _are_ using it correctly --

Quite.

>You do have one sub-element in
there that points to a 3rd party search engine, and that's pointed
specifically at your site. You're in a grey area here: _Should_ such
3rd party searches fall under the remit of what OpenSearch describes?,

Why shouldn't they? It's a search service dedicated to one site. What
does it matter that the server/ domain that it runs on is not the same;
and how does that differ from large organisation, exmaple.com, whose
search runs on a dedicated sever at search.examle.com or
example-search.com??

>but thinking about yours more carefully, I now think you're on the
right side of things.

This isn't a _substantial_ use of OpenSearch and it's still up for
debate if this is the best was to do it -- but it's not _wrong_ to do
it this way.

>Well, that's certainly one opinion, but it's not my reading of either:

It's still very much my reading of it.

You have failed to quote my quote of your claim:

>>This would appear to be a fundamental mis-use of the purpose of
OpenSearch

>OpenSearch is only there to
describe dynamic search engines and it's a mis-use of it to use it to
merely describe static site metadata.

Who is merely describing static metadata?

> http://www.last-child.com/add-opense...your-web-site/

"create an XML document that defines your site's search engine
methods"

Which is exactly what I've done.
Now, would someone with IE7 kindly check out the "opensearch.xml" file
referenced from:

http://www.westmidlandbirdclub.com/new.htm

for me, please?
--
Andy Mabbett
Say "NO!" to compulsory ID Cards: <http://www.no2id.net/>

Free Our Data: <http://www.freeourdata.org.uk>

[Andy Dingley]

Andy Mabbett wrote:

Now, would someone with IE7 kindly check out the "opensearch.xml" file
referenced from:

http://www.westmidlandbirdclub.com/new.htm

IE stuff:

IE7 doesn't support OpenSearch (usefully). M$'s current paranoia has
broken it - you aren't permitted to access this terrible security hole
unless you mark the site as trusted.

Using OpenSearch for the first time with a site is too long-winded to
be useful. You have to add the sie provider, then submit the search. No
one-click searching, which is ridiculous.

You don't even see that a site supports OpenSearch unless you know
where to look.

IE doesn't see any connection between OpenSearch and the current page.
You have to pick a search provider from a list, you aren't
automatically defaulted to one because you're already on the relevant
page.

Site stuff:

Your shortname of "WMBC" is possibly too short to be meaningful in a
list of providers. Remember that IE is enforcing a two-step "register,
then search" approach here, so anyone who uses OpenSearch even casually
is likely to collect a long list of search providers.

The results are in HTML, not anything aggregateable. This kills most of
the interesting uses of OpenSearch.

Your <tagsshould probably include "birdwatching" and as many related
terms you can squeeze in there.

General stuff.

OpenSearch isn't about replacing a site's search box. Certainly not in
the short term, as usability-wise it's substantially les usable than
that.

What OpenSearch is really for is for syndicating searches across
multiple related search providers, typically per-site searches for
related sites. To make this work we need <tags(sic) in the search
metadata to indicate "related" search providers, and we need search
results to be returned in a format that's not just easily parseable,
but that is _identically_ parseable (i.e. a standard format like Atom)
without needing to write site-specific parsers.

OpenSearch isn't well designed (IMHO). In particular the <tagselement
is almost unusable. This really did need extensible scope for _much_
smarter annotation, particularly structured annotation and the ability
to use shared vocabularies (e.g. TGN for your geographical location).
As it is, the OpenSearch "trivial lost of keywords" is barely useful
here, and we've lost the possibility of auto-aggregation by
auto-discovery of relevant search providers.

I've spent most of the last decade faffing about with metadata - to
have Amazon produce a protocol so trivial and restricted this late in
the game was a little galling. We _know_ how to do it better, and we
even know how to do it in a simplicity-compatible manner. Shame they
didn't pay attention.

[Andy Dingley]

Andy Dingley wrote:

What OpenSearch is really for is for syndicating searches across
multiple related search providers, typically per-site searches for
related sites. To make this work we need <tags(sic) in the search
metadata to indicate "related" search providers, and we need search
results to be returned in a format that's not just easily parseable,
but that is _identically_ parseable (i.e. a standard format like Atom)
without needing to write site-specific parsers.

In reply to an email (keep it on Usenet, guys!)

The history of OpenSearch is clear. Amazon invented it to _feed_ their
A9 engine, then M$ adopted it later as a description format for adding
search engines to IE7. This was a smart move, as it avoided them having
to invent Yet Another Protocol, but it also made the fundamental reason
behind OpenSearch less obvious. If your OpenSearch "compliant" engine
only supports HTML results, then you're serving the human(sic) IE7
users, but you're ignoring the real potential of OpenSearch as a format
for syndicating searches.

Z39.50 was intended as a real-time search aggregator, where distributed
search engines would be queried in response to a central human query.
OpenSearch is a little different to that, in that it's clearly intended
to support search as an automated knowledge discovery process - i.e.
rather than just "importing published metadata" (on topics that the
providers choose), centralised search engines can "ask around" for
topics that they're interested in enquiring about. They can do this
before any immediate human queries, as a means of expanding their
knowledge bases.
Jon Udell on the first appearance
http://weblog.infoworld.com/udell/2005/03/28.html

M$ on its adoption
http://blogs.msdn.com/ie/archive/2005/09/14/466278.aspx