[p2pu-dev] first impressions on new.p2pu

Paul Osman paul at mozillafoundation.org
Wed May 25 01:51:06 UTC 2011

I haven't used either, but I know that the AMO team (as well as others 
at Mozilla) have moved away from Sphinx in favour of ElasticSearch 
(another Java / Lucene based search engine). ES has been in use for a 
long while by our metrics team who push a ton of data through it. I 
can't recall the exact reasons why ES was chosen over Solr, but the move 
away from sphinx was prompted by a feeling that Sphinx was too 
complicated and hard to support.


On 11-05-23 09:38 AM, Jessy Kate Schingler wrote:
> Cool, I'm not familiar with sphinx search. I wonder if Paul has any
> thoughts/experience on sphinx search versus solr? Adding him to cc :)
> Jessy
> On Monday, May 23, 2011, zuzel.vp<zuzel.vp at gmail.com>  wrote:
>> On Mon, May 23, 2011 at 7:16 AM, Jessy Kate Schingler
>> <jessy at jessykate.com>  wrote:
>>> re user sign up-- sorry i wasn't clear. i meant signing up for a course, not
>>> an account. he already had an account, but couldn't find where to sign up
>>> for a specific course that was open for sign up. for these courses, could we
>>> have a bigger button(s) near the top of the course homepage?
>> http://p2pu.lighthouseapp.com/projects/71002/tickets/337-move-follow-participate-and-contact-organiser-buttons-above-the-fold
>>> re search, have you thought about something like
>>> solr? http://lucene.apache.org/solr/
>>> solr indexes along any fields you define (typically multiple), to enable
>>> fast real time search given a large variety of potential parameters/filters.
>>> it trades off space (ie, the indexes can get big) for speed and scalability,
>>> basically. it's used on many "high traffic
>>> websites": http://wiki.apache.org/solr/PublicServers.
>>> i set up solr for one of my projects at sunlight and worked a lot with index
>>> definitions; if there's interest i would be happy to set it up for us. it's
>>> free and open source. index space isn't an issue for us AFAIK, but the main
>>> potential drawback i could imagine is that our dev server probably couldn't
>>> handle the extra load. thoughts?
>> Any experiments on implementing search are welcome, and now that we
>> can import course content from the old site it will be easier to test
>> performance in a dev instance. However, I recommend not including it
>> in production until we are sure both search and indexing works for
>> larges amounts of data (at least comparable to what we have in the old
>> site right now), and that it does not cause lock errors.
>> The last time we meet with Paul we talked about
>> http://sphinxsearch.com/ with is used by
>> https://github.com/jbalogh/zamboni/tree/master/apps/search. That is
>> another one we could experiment with, though it also has its
>> performance issues.
>> --
>> Thanks,
>>      Zuzel
>>> Jessy
>>> --
>>> http://jessykate.com
>>> On Mon, May 23, 2011 at 1:35 AM, Philipp Schmidt<phi.schmidt at gmail.com>
>>> wrote:
>>>> Search is a problem we hoped batucada would solve.
>>>> Zuzel: could you check with Paul what the status on that is?
>>>> To reduce confusion on the courses list page, we could also consider
>>>> listing draft courses on a separate page.
>>>> P
>>>> On 23 May 2011 12:17 AM, "zuzel.vp"<zuzel.vp at gmail.com>  wrote:
>>>>> Sign up/Registration of new accounts was temporarily hidden on
>>>>> purpose. Before the migration occurs we need users not to create new
>>>>> accounts unless they had never used the old site so they keep their
>>>>> identities. After the creation of new accounts in the old site is
>>>>> disabled we will migrate all user accounts to the new site. At that
>>>>> point we will not need anymore the current restriction of not allowing
>>>>> any user to register accounts with username/email from the old site
>>>>> (unless they proved that they are the same person by login in with
>>>>> their username and password from the old site). After the migration we
>>>>> can make signup more visible again (either by having all the forms on
>>>>> the same page or by adding a direct link like in drumbeat/batucada).
>>>>> Most of the tickets are not assigned to a milestone now because we had
>>>>> to re-schedule things in order to move the migration to an early date.
>>>>> However even before this, search was assigned I think to milestone 0.9
>>>>> because it is a feature hard to implement in a way that scales. All
>>>>> solutions I have seen including search in a django project tend to
>>>>> become considerable slow when they are setup in a real world scenario.
>>>>> One solution could be to rely google for including search capabilities
>>>>> (http://www.google.com/cse/).
>>>>> Thoughs?
>>>>> --
>>>>> Thanks,
>>>>> Zuzel
>>>>> On Sun, May 22, 2011 at 5:13 PM, Alison Jean Cole
>>>>> <alisonjean.cole at gmail.com>  wrote:
>>>>>> Agree with both of Jessy's suggestions.
>>>>>> ALISON
>>>>>> new.p2pu.org/en/alison/
>>>>>> On Sat, May 21, 2011 at 10:48 AM, John Britton
>>>>>> <public at johndbritton.com>
>>>>>> wrote:
>>>>>>> I also think we should make the signup/signin box happen on the same
>>>>>>> page
>>>>>>> as opposed to redirecting to a new page, less friction overall.
>>>>>>> --
>>>>>>> contact info:
>>>>>>> http://www.johndbritton.com
>>>>>>> @johndbritton - http://twitter.com/johndbritton
>>>>>>> On Sat, May 21, 2011 at 1:43 PM, Jessy Kate Schingler
>>>>>>> <jessy at jessykate.com>  wrote:
>>>>>>>> 2 pieces of feedback on someone looking at the new.p2pu site.
>>>>>>>> i had recommended they create an account in order to sign up for a
>>>>>>>> specific course. so of course the first thing they looked for was
>>>>>>>> a search
>>>>>>>> box. there's quite a lot of courses now, so i wonder if the "search
>>>>>>>> box"
>>>>>>>> ticket on lighthouse might deserve to be moved up a bit in priority
>>>>>>>> (currently "unscheduled")?

More information about the p2pu-dev mailing list