[Search-l] Browser Test

[seth]

K...so here's a response from somebody who does have a more official  
connection to the project.

First, I'm very sorry the search results page does not work for you.   
I'll bug jer and some other wikia people later today and see if we  
can't replicate the problem and resolve it.  I think I managed to  
scrape together that you're using IE5, if you could send any  
additional details of firewall/NAT/PC configuration (even to me  
personally) it would be appreciated.  Also, if you have some skills  
with HTML/CSS/JavaScript feel free to check out the source for 'the  
search page' from our svn repository at http://svn.swlabs.org/ 
re.search/.  Also, some have sad they wish to make their own  
interfaces, feel free to use the SVN sources as a reference on how to  
query the components.

The 'search results' page pulls data from three primary sources, all  
using JSON (http://www.json.org/) requests.  One source is the wikia  
mini-article, the 2nd the 'people search' from the social platform  
(foowi: http://svn.swlabs.org/foowi/), the third is of course the  
search engine itself.

As much as the search page is broken for you, part of deciding to  
have the browser combine these three systems client side was to  
hopefully increase the perceived availability of the system.  By  
having the web browser do the combination of the results if any one  
of our three 'data sources' go offline, only that component of the  
web page does not load.  The others will load just as quickly as they  
normally would, despite the partial outage.

The quickness of loading the page, or at least as much as quickly as  
possible, was one reason for choosing to build the results page this  
way.  With a server combining the outputs of these three systems it  
would have to wait for the slowest system to respond (or time out)  
before showing you anything.  And, with the unknown amount of traffic  
that was going to hit us day one, we were expecting at least one of  
those major subsystems to go offline.  We didn't know which would  
succumb, and luckily none did, but if any had this system would have  
degraded quite nicely.

As for if we ever plan to offer a pure 'plain' HTML interface...I  
don't know.  I'm far from a person who sets direction for this kind  
of project, and right now we're focusing on things that can improve  
quality of search results.

Hope that explains things a little.

seth


On Feb 12, 2008, at 9:25 AM, Tom Wright wrote:

> Just a warning - I'm in no way an official source of information.
>
> The point isn't so much that the webservers are separate machines but
> that they are on a seperate LAN, so for each external querie you have
> to open a connection outside the lan - But this is just conjecture on
> my part.
>
> Yes there is an openly accessible api. (Though I'm not sure whether
> it's syntax is permanently fixed yet.) The URL is:
>
> http://not-quite-ready-yet.index.swlabs.org/nutchsearch? 
> query=<<QUERY>>&hitsPerSite=1&lang=en&hitsPerPage=10&type=json&start=< 
> <START
> DISPLAY FROM>>.
>
> This is talking to a nutch interface - so if you are trying to play
> with parameters looking at the nutch documentation might help.
>
> If you implement a serverside front-end, it might be useful if you'd
> submit the code back to search wikia. This could be used a fallback
> front-end if javascript is not supported.
>
> Remember (as I'm sure you've heard several times) that search wikia is
> still a work in progress, so some things will likely be slightly
> broken.
>
> Tom Wright
>
>
>
>
> On Feb 12, 2008 2:53 PM, Markus Petz <markus.petz at chello.at> wrote:
>> At 12.02.2008 15:02, you wrote:
>>> The use of javascript is fundamental to the current approach - it is
>>> not just being used for presentation or dynamic effects.
>>
>> Thank you for the expanation. It makes me think you wanted too much
>> functionality whilst not being able to provide the basics. You are  
>> going to
>> have a lot of troubles if you rely on client-side presentation  
>> only. There
>> should be at least the option to have a clean server-provided  
>> presentation
>> page. This has nothing to do with the database server setup. you  
>> can still
>> have any db setup you like, with all the remote access and stuff. The
>> webserver(s) naturally are separate machines. As a user i could  
>> care less
>> how it works, if it works.
>>
>> That does mean there will be an openly accessible API to the  
>> search server?
>> In that case i might be wanting to write my own quick frontend on  
>> my own
>> HTTP server, bypassing all the unnecessary stuff of wikia.
>>
>>
>> _______________________________________________
>> Wikia Search mailing list
>> http://alpha.search.wikia.com/
>> Change options or unsubscribe: http://lists.wikia.com/mailman/ 
>> options/search-l
>>
> _______________________________________________
> Wikia Search mailing list
> http://alpha.search.wikia.com/
> Change options or unsubscribe: http://lists.wikia.com/mailman/ 
> options/search-l