On 11 March 2016 at 10:52, ViswaPrabha (വിശ്വപ്രഭ) vp2007@gmail.com wrote:
Failed my dream :(
Any string in any language in any wikipedia project. How far is my dream?
I share your dream! :-)
Unfortunately, the dream is quite far away from reality. Querying every search index would put a big performance strain on the search servers. Additionally, it would likely return you a bunch of really irrelevant results, so there's a lot of user experience implications that would need to be figured out as well. Discovery is not actively working on this at present.
Thanks, Dan
Lovely work.
On Mar 11, 2016 5:32 PM, "Dan Garry" dgarry@wikimedia.org wrote:
On 11 March 2016 at 10:52, ViswaPrabha (വിശ്വപ്രഭ) vp2007@gmail.com
wrote:
Failed my dream :(
Any string in any language in any wikipedia project. How far is my
dream?
I share your dream! :-)
Unfortunately, the dream is quite far away from reality. Querying every search index would put a big performance strain on the search servers.
Perhaps you could do this w two queries, one to a composite index that is only updated weekly.
Additionally, it would likely return you a bunch of really irrelevant
results,
Make this opt-in, add a different background color for results from the all-language index, & divide their search-relevance by a language-prominence factor...
+SJ
On 11 March 2016 at 16:09, Samuel Klein meta.sj@gmail.com wrote:
Perhaps you could do this w two queries, one to a composite index that is only updated weekly.
Indeed, there are mechanisms that can make cross-wiki searching more feasible. In fact, one mechanism we are in the very early stages of exploring is merging all the projects in a given language into a single index, so that one could search *all* projects in a language, rather than just a single project in a given language. I have no timeline here; as I said, we're in the very early stages, and of course we have other work on the go at the same time.
Additionally, it would likely return you a bunch of really irrelevant
results,
Make this opt-in, add a different background color for results from the all-language index, & divide their search-relevance by a language-prominence factor...
I plan to worry more about the user experience implications that I mentioned once we're a bit closer to solving the technical feasibility questions. As you've shown, these are definitely solvable problems, but I don't want to put the cart before the horse, as it were. :-)
Thanks, Dan
On Mar 11, 2016 19:32, "Dan Garry" dgarry@wikimedia.org wrote:
I plan to worry more about the user experience implications that I mentioned once we're a bit closer to solving the technical feasibility questions.
Fair enough! Great once more to see this update. It makes this a much more engaging page to set as a browser default on public machines.
Sj
On 12.03.2016 01:09, Samuel Klein wrote:
Lovely work.
On Mar 11, 2016 5:32 PM, "Dan Garry" dgarry@wikimedia.org wrote:
On 11 March 2016 at 10:52, ViswaPrabha (വിശ്വപ്രഭ) vp2007@gmail.com
wrote:
Failed my dream :(
Any string in any language in any wikipedia project. How far is my
dream?
I share your dream! :-)
Unfortunately, the dream is quite far away from reality. Querying every search index would put a big performance strain on the search servers.
Perhaps you could do this w two queries, one to a composite index that is only updated weekly.
Additionally, it would likely return you a bunch of really irrelevant
results,
Make this opt-in, add a different background color for results from the all-language index, & divide their search-relevance by a language-prominence factor...
Which ideally can be taken from the clients "accept language" request header.
Purodha
wikitech-l@lists.wikimedia.org