From jwales at wikia.com Fri Aug 1 03:08:19 2008 From: jwales at wikia.com (Jimmy Wales) Date: Thu, 31 Jul 2008 20:08:19 -0700 Subject: [search-ui] [Search-l] Widgets In-Reply-To: <24a809760807291909kab5f7acl79723095cb1c7c0e@mail.gmail.com> References: <4884B486.5040405@wikia.com> <488622F9.3050307@wikia.com> <48870619.7050608@gmx.de> <24a809760807231642h316b9cd2vcacb450e626aaf2e@mail.gmail.com> <488A305A.6010405@wikia.com> <24a809760807291909kab5f7acl79723095cb1c7c0e@mail.gmail.com> Message-ID: <48927E23.8030905@wikia.com> I have no opposition to "searchlets" except to note that they are mostly going to be useless for most end users. To use a "searchlet" like "now:" or "twitter:" or "torrent:" users have to learn something; most will not. So although I have no opposition, I am not very excited by them. What I am excited by is the application framework that invites partners to help us improve search by submitting regular expressions that trigger the return of useful objects. If I type "jimmy wales twitter"... or rather anything that matches: (.*?) twitter then I should get a widget that directs me to the 'searchlet'. Did you want to search twitter for "jimmy wales"? Where "jimmy wales" is a link to summize.com?query=%1; or whatever it would be. --Jimbo adatta02 at gmail.com wrote: > We've been playing around with "widgets" and other "injections" for the > last few days. > > So far, we've broken down search manipulations into "searchlets" and > "widgets". > > As Aaron put it, "searchlets" consist of searches that support searching > within some subset of the universal search space. We've been defining > these using a ":" syntax. An example of this search is > http://fp029.sjc.wikia-inc.com/search/search.html#twitter:earthquake > which uses summize to search twitter. We also have "video", "users", and > "torrent" built which presents results in a similar fashion (in the > normal search view with KT accessible). > > Other "searchlets" we were throwing around are "now:" which would query > the "nowsphere" (twitter, latest blog posts, ect) to try and get a sense > of what people are saying about something right now. The vague aspects > of "searchlets" seem to be presentation and how they interact with KT. > > A big question is should searchlets display in a different layout than > regular search (ie should we be able to display videos in a table > instead of a flat list? > http://fp029.sjc.wikia-inc.com/search/search.html#video:wikia ) > > The KT question is basically what happens in the corner cases. For > example, on a resultset for twitter:earthquake if a user "stars" or > "spotlights" an item and then the item fall off the first page (twitter > searches are sorted by date). Should this item automatically promote to > the first page or should the KT actions be ignored? > > We've also been playing around with how to impliment and display > "widgets" but haven't come to much of a consensus. So far, we've been > "defining" widgets as additional discrete pieces of information (on top > of the nutch/KT results). > > Jeff's implimented several widgets directly in our JS that display > in-line with the search results. For example, > http://fp029.sjc.wikia-inc.com/search/search.html#traffic%2010003 > renders traffic information as a normal search result. > > We also experimented with rendering the widgets on top of the results - > http://fp029.sjc.wikia-inc.com/search/search.html#Newark,%20NJ > > The issue then is, should widgets render in-line with the other results > and therefore be subject to the same KT operations (star, delete, > spotlight, ect) or should they render in a seperate area and be > unaffected by KT. Also, what would the behavior be for keywords that > return multiple widgets (ie weather 10003) > > Ashish > > On Fri, Jul 25, 2008 at 3:58 PM, Jimmy Wales > wrote: > > This is very interesting. :) > > adatta02 at gmail.com wrote: > > The big concern with allowing server side code is security. From the > > chat, it sounds like the plan is to hand review any code that will be > > run on the server and reject "dangerous" code. This seems like a > > daunting if not down right impossible task as any reasonably > complicated > > piece of code is going to be extremely hard to certify as "safe". > There > > are various examples of sneaking nefarious code into a seemingly safe > > code block famously > http://en.wikipedia.org/wiki/Underhanded_C_Contest > > Well, this is a typical problem faced by open source projects all the > time. Whom do we trust to "commit"? I think we just need to have clear > terms and be careful. Excessive paranoia is not warranted. > > > From a developer's standpoint - the big gain of being able to run > > server side code is the ability to interact with the OS layer (files, > > sockets, databases, ect) and libraries (GD, standard libraries, > ect) . > > However, it seems like there is no way we would be able to provide > > access to any OS features while simultanously gaurenteeing > security. It > > also seems like providing persistent storage through this framework > > would require significant infastructure. > > I don't think that's such a big part of it. I would say that for both > us and for developers, the major gain of running server side is > performance and integration. > > I suppose there are 3 possibilities: > > 1. "pure" client side - javascript can compute 3*7 and return 21. No > risk, all is well. EXAMPLE: calculator > > 2. "pure" server side - our servers compute something complex and > relying on databases, etc. - one huge advantage of doing this is that it > makes the code all open-source > > 3. third-party "RESTful" calls - no risk, but performance is an issue, > as is proprietary-ness EXAMPLE: weather > > > What I'd propose is to construct a framework that mirrors the Google > > igoogle system. Following from this, developers would be able to > write > > javascript blocks that run on some set of triggers (either regex > matches > > or simple in_string()) Next, on the search results page, the user's > > browser executes this javascript snippit in an iframe (to create a > > sandbox) and injects any new HTML into the actual search results. > > I have no opposition to this, except for the possible performance > issue... we can require that before we send out the javascript, it has > to be open source. > > > I think allowing javascript snippits would provide a good mix between > > ease of development, security, as well as robustness. A javascript > > widget would be able to do anything that a server side widget could > > accomplish. It would be able to fetch and process RSS feeds, make > cross > > site XHR requests, and leverage additional "enviroment" variables we > > choose to make available to it. > > I am nearly convinced. > > > > > _______________________________________________ > Search-UI mailing list > Search-UI at wikia.com > http://lists.wikia.com/mailman/listinfo/search-ui > > > > ------------------------------------------------------------------------ > > _______________________________________________ > Search-UI mailing list > Search-UI at wikia.com > http://lists.wikia.com/mailman/listinfo/search-ui From balinny at gmail.com Fri Aug 1 18:36:03 2008 From: balinny at gmail.com (Balinny) Date: Fri, 01 Aug 2008 20:36:03 +0200 Subject: [search-ui] [Search-l] Widgets In-Reply-To: <9131eace0807292131ye888d3fv69c7475c878cdb7c@mail.gmail.com> References: <488622F9.3050307@wikia.com> <48870619.7050608@gmx.de> <24a809760807231642h316b9cd2vcacb450e626aaf2e@mail.gmail.com> <488A305A.6010405@wikia.com> <24a809760807291909kab5f7acl79723095cb1c7c0e@mail.gmail.com> <9131eace0807292131ye888d3fv69c7475c878cdb7c@mail.gmail.com> Message-ID: <48935793.4040904@gmail.com> Jeff wrote: > One problem that this presents, however, is with widgets that are > providing time sensitive / news-type data. For example, if the > weather widget is editable... I think that the solution is making the edit interface part of the widget. So, a widget must have a set of apis, such as show, voted, edit... When you ask to edit the result, a weather widget would refuse it (typically, it would have hidden the edit link), a wiki widget could send you to the edit page, and a map widget would allow you to move the position in the map. Jimmy Wales wrote: > I have no opposition to "searchlets" except to note that they are mostly > going to be useless for most end users. To use a "searchlet" like "now:" > or "twitter:" or "torrent:" users have to learn something; most will not. > Add a set of option boxes below the textbox, like the ones Google have for the web / pages from site / pages from country, but offering: Search in the nowsphere, twitter changes, videos, users, More With more opening a popup with the full list. I'd show inlnie the last used by the user, the most popular ones (eg. images and videos) and some of them at random, so they can learn about the new ones. People can learn that using it just prefixes the query with the searchlet, or just click it each time :) From awgibbs at awgibbs.com Sat Aug 2 07:07:18 2008 From: awgibbs at awgibbs.com (Andrew W. Gibbs) Date: Sat, 2 Aug 2008 03:07:18 -0400 Subject: [search-ui] sliding window for infinite scrolling Message-ID: <20080802070718.GB30297@raptor.commandosoftware.com> The infinite scrolling is nifty and novel from a usability standpoint. I wonder, however, whether it might cause resource utilization problems since the earlier items continue to accumulate as you scroll. This may seem inconsequential from the standpoint of a desktop machine, but consider usage on a mobile device. Where as the conventional paging style guarantees O(1) memory usage, the infinite scrolling style is O(n). To preserve the general UI mechanics but attain O(1) memory usage, you might instead have a sliding window of search results with things falling off the top or bottom if the user scrolled backward or forward respectively. The challenge, of course, would be that you could no longer free-ride on the browser's native GUI faculties. As it stands, you can "cheat" when scrolling downward because the end of the search results coincides with the bottom of the page, but you don't have this luxury at the top of the page. You could theoretically deal with that with frames, but frames feel oh so 1996. -- AWG -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://lists.wikia.com/pipermail/search-ui/attachments/20080802/b97f572a/attachment.bin From newsmarkie at googlemail.com Mon Aug 4 16:39:07 2008 From: newsmarkie at googlemail.com (Mark (Markie)) Date: Mon, 4 Aug 2008 17:39:07 +0100 Subject: [search-ui] i18n errors In-Reply-To: <488274F3.9030002@gmx.de> References: <488274F3.9030002@gmx.de> Message-ID: bump, still like it.... regards mark On Sun, Jul 20, 2008 at 12:12 AM, Rainer Blome wrote: > Mark (Markie) wrote: > > http://c.imagehost.org/0228/untitled_4.jpg button is borked. > Fixed, but not ready for public consumption. > Jer, you offered a SVN account with checkin capability. > Does this offer still stand? > That would allow for individual fixes independent of bigger > transformations. > > but i am loving the on the fly translations :-) > The breakage makes Jeff's technique used to implement on the fly > translation visible :-). > In this case, it is incorrectly applied. > > Rainer > > _______________________________________________ > Search-UI mailing list > Search-UI at wikia.com > http://lists.wikia.com/mailman/listinfo/search-ui > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.wikia.com/pipermail/search-ui/attachments/20080804/765e4279/attachment.html From rainer.blome at gmx.de Wed Aug 6 00:46:09 2008 From: rainer.blome at gmx.de (Rainer Blome) Date: Wed, 06 Aug 2008 02:46:09 +0200 Subject: [search-ui] i18n errors In-Reply-To: References: <488274F3.9030002@gmx.de> Message-ID: <4898F451.70005@gmx.de> Mark (Markie) wrote: > bump, still like it.... > On Sun, Jul 20, 2008 at 12:12 AM, Rainer Blome > wrote: > Mark (Markie) wrote: > > http://c.imagehost.org/0228/untitled_4.jpg button is borked. > Fixed, Checked in a fix for this particular button. Please let me know of any other annoyingly visible HTML. > but not ready for public consumption. The current idea for the general fix is to use separate "gettext" functions for different usage types: o get_text(key, args) - fetches the text (format string) in i18n.onfly_lang[key] and uses it and the given args to format the translated text. Records untranslated texts. Used by updater functions and all other translation functions. o get_span(key, args) - returns a span with a generated id whose inner HTML is the translated text (whatever get_text returns). As a side effect, registers an updater function. Like Jeffrey's gettext mod when called with*out* args. o get_title(key, args, param) - returns a span with id=param.id whose inner HTML is the translated text (whatever get_text returns). As a side effect, registers an updater function. Like Jeffrey's gettext mod when called *with* args. {id:sel.name, param:'options[' + i + '].text', pre:o.count+" "} This might be split up further. This still needs work, though. Rainer From jeremie at jabber.org Wed Aug 6 22:02:00 2008 From: jeremie at jabber.org (Jeremie Miller) Date: Wed, 6 Aug 2008 17:02:00 -0500 Subject: [search-ui] sliding window for infinite scrolling In-Reply-To: <20080802070718.GB30297@raptor.commandosoftware.com> References: <20080802070718.GB30297@raptor.commandosoftware.com> Message-ID: I'd prefer to start breaking out the UI for different types/ experiences, in subversion I checked in an exp directory with a mostly stripped down version where it's easier to do experimenting with UI. I even started working on an iphone-specific version ages ago, I'm more for seeing lots of these skins than packing everything into one single interface :) On Aug 2, 2008, at 2:07 AM, Andrew W. Gibbs wrote: > The infinite scrolling is nifty and novel from a usability standpoint. > I wonder, however, whether it might cause resource utilization > problems since the earlier items continue to accumulate as you scroll. > This may seem inconsequential from the standpoint of a desktop > machine, but consider usage on a mobile device. Where as the > conventional paging style guarantees O(1) memory usage, the infinite > scrolling style is O(n). To preserve the general UI mechanics but > attain O(1) memory usage, you might instead have a sliding window of > search results with things falling off the top or bottom if the user > scrolled backward or forward respectively. The challenge, of course, > would be that you could no longer free-ride on the browser's native > GUI faculties. As it stands, you can "cheat" when scrolling downward > because the end of the search results coincides with the bottom of the > page, but you don't have this luxury at the top of the page. You > could theoretically deal with that with frames, but frames feel oh so > 1996. > > -- AWG > _______________________________________________ > Search-UI mailing list > Search-UI at wikia.com > http://lists.wikia.com/mailman/listinfo/search-ui From phil.nelson at gmail.com Wed Aug 6 23:02:15 2008 From: phil.nelson at gmail.com (Christian Nelson) Date: Wed, 6 Aug 2008 19:02:15 -0400 Subject: [search-ui] Allow users to select multiple languages Message-ID: I saw this idea on this blog: http://bani2.blogspot.com/2008/08/wikia-evolution.html To my knowledge, there's no good interface in any of the big search engines that allows users who speak more than one language to select those languages and have them apply to the same search. E.G. returning results in english AND french, not just english OR french. This is something that I think could help set Wikia apart, especially with the large support other wiki-style projects seem to get in many non-US countries. - Phil Nelson http://philnelson.name From jeremie at jabber.org Wed Aug 6 23:06:20 2008 From: jeremie at jabber.org (Jeremie Miller) Date: Wed, 6 Aug 2008 18:06:20 -0500 Subject: [search-ui] Allow users to select multiple languages In-Reply-To: References: Message-ID: <596B329F-9123-4A87-A0AA-909BC86C321E@jabber.org> That's a pretty wicked idea, should be totally possible with the framework, but will take some extra JS logic to merge everything together... definitely on the list IMO, now I just need to fluently learn another language :) Jer On Aug 6, 2008, at 6:02 PM, Christian Nelson wrote: > I saw this idea on this blog: http://bani2.blogspot.com/2008/08/wikia-evolution.html > > To my knowledge, there's no good interface in any of the big search > engines that allows users who speak more than one language to select > those languages and have them apply to the same search. E.G. returning > results in english AND french, not just english OR french. This is > something that I think could help set Wikia apart, especially with the > large support other wiki-style projects seem to get in many non-US > countries. > > - Phil Nelson > http://philnelson.name > _______________________________________________ > Search-UI mailing list > Search-UI at wikia.com > http://lists.wikia.com/mailman/listinfo/search-ui > From jwales at wikia.com Wed Aug 6 23:32:50 2008 From: jwales at wikia.com (Jimmy Wales) Date: Wed, 06 Aug 2008 19:32:50 -0400 Subject: [search-ui] [Search-l] Widgets In-Reply-To: <48935793.4040904@gmail.com> References: <488622F9.3050307@wikia.com> <48870619.7050608@gmx.de> <24a809760807231642h316b9cd2vcacb450e626aaf2e@mail.gmail.com> <488A305A.6010405@wikia.com> <24a809760807291909kab5f7acl79723095cb1c7c0e@mail.gmail.com> <9131eace0807292131ye888d3fv69c7475c878cdb7c@mail.gmail.com> <48935793.4040904@gmail.com> Message-ID: <489A34A2.5080806@wikia.com> Balinny wrote: > Jeff wrote: >> One problem that this presents, however, is with widgets that are >> providing time sensitive / news-type data. For example, if the >> weather widget is editable... > I think that the solution is making the edit interface part of the > widget. So, a widget must have a set of apis, such as show, voted, edit... > When you ask to edit the result, a weather widget would refuse it > (typically, it would have hidden the edit link), a wiki widget could > send you to the edit page, and a map widget would allow you to move the > position in the map. That sounds right to me. Also, there is the possibility of editing the "template" rather than the actual result. For example... let's suppose we have a "famous art works" widget. Someone creates a huge list of famous artworks by title. When triggered, it gives back a thumbnail of that artwork, a short snippet from Wikipedia, and links to 3-5 major art websites with quality content. Niiiiice. And then, a new major art site comes online. Someone notices it and wants to add it... not by going to 50,000 famous artworks and editing the item one by one, but by editing the *template*. (This probably necessitates that the website in question needs to have a url structure which would allow predictable linking... but that's their job!) --Jimbo From jeremie at jabber.org Tue Aug 12 08:20:48 2008 From: jeremie at jabber.org (Jeremie Miller) Date: Tue, 12 Aug 2008 03:20:48 -0500 Subject: [search-ui] some experimental bits Message-ID: <4AF0F24C-FB5D-49FE-96B9-91E892F5AE55@jabber.org> Played with an accordion style result, spent too much time just finding the js to do it and at then not enough time making it work smoothly, but you'll get the general idea: http://re.search.swlabs.org/BEWARE2/exp/accordion.html Updated the "light" experiment by simply doing a float left, resize your browser and watch it all reflow, is really intriguing IMO: http://re.search.swlabs.org/BEWARE2/exp/light.html#apple Jer From aaron.wright at gmail.com Tue Aug 12 08:29:54 2008 From: aaron.wright at gmail.com (Aaron Wright) Date: Tue, 12 Aug 2008 04:29:54 -0400 Subject: [search-ui] some experimental bits In-Reply-To: <4AF0F24C-FB5D-49FE-96B9-91E892F5AE55@jabber.org> References: <4AF0F24C-FB5D-49FE-96B9-91E892F5AE55@jabber.org> Message-ID: I really like the multiple column results. Maybe we could make an option to allow users to toggle between single and multiple columns. --Aaron On Tue, Aug 12, 2008 at 4:20 AM, Jeremie Miller wrote: > Played with an accordion style result, spent too much time just > finding the js to do it and at then not enough time making it work > smoothly, but you'll get the general idea: > http://re.search.swlabs.org/BEWARE2/exp/accordion.html > > Updated the "light" experiment by simply doing a float left, resize > your browser and watch it all reflow, is really intriguing IMO: > http://re.search.swlabs.org/BEWARE2/exp/light.html#apple > > Jer > > _______________________________________________ > Search-UI mailing list > Search-UI at wikia.com > http://lists.wikia.com/mailman/listinfo/search-ui > From balinny at gmail.com Tue Aug 12 14:31:30 2008 From: balinny at gmail.com (Balinny) Date: Tue, 12 Aug 2008 16:31:30 +0200 Subject: [search-ui] some experimental bits In-Reply-To: <4AF0F24C-FB5D-49FE-96B9-91E892F5AE55@jabber.org> References: <4AF0F24C-FB5D-49FE-96B9-91E892F5AE55@jabber.org> Message-ID: <48A19EC2.8040902@gmail.com> Jeremie Miller wrote: > Played with an accordion style result, spent too much time just > finding the js to do it and at then not enough time making it work > smoothly, but you'll get the general idea: > http://re.search.swlabs.org/BEWARE2/exp/accordion.html > I have a doubt. How do you visit the results there? There's a link to the page in the title (i see it on the status bar) but click on it just folds/unfolds the description... From jeremie at jabber.org Tue Aug 12 17:21:46 2008 From: jeremie at jabber.org (Jeremie Miller) Date: Tue, 12 Aug 2008 12:21:46 -0500 Subject: [search-ui] some experimental bits In-Reply-To: <48A19EC2.8040902@gmail.com> References: <4AF0F24C-FB5D-49FE-96B9-91E892F5AE55@jabber.org> <48A19EC2.8040902@gmail.com> Message-ID: >> http://re.search.swlabs.org/BEWARE2/exp/accordion.html > I have a doubt. How do you visit the results there? There's a link to > the page in the title (i see it on the status bar) but click on it > just > folds/unfolds the description... Yeah, it's just an experiment, wasn't sure if it was worth cleaning up and making usable, it's open source so I suppose if someone likes it they've got some obvious things to fix right away :) Jer From jeremie at jabber.org Tue Aug 12 17:23:57 2008 From: jeremie at jabber.org (Jeremie Miller) Date: Tue, 12 Aug 2008 12:23:57 -0500 Subject: [search-ui] some experimental bits In-Reply-To: References: <4AF0F24C-FB5D-49FE-96B9-91E892F5AE55@jabber.org> Message-ID: > I really like the multiple column results. Maybe we could make an > option to allow users to toggle between single and multiple columns. I do as well, and if we figure out a better way to show the mouseover menu (which takes up a lot of margin space) then yeah it's just some CSS so it could be fairly easily integrated into the current result UIs. Jer From rainer.blome at gmx.de Sat Aug 16 21:12:55 2008 From: rainer.blome at gmx.de (Rainer Blome) Date: Sat, 16 Aug 2008 23:12:55 +0200 Subject: [search-ui] HTML comment instead of newline at end of .html files? Message-ID: <48A742D7.4010303@gmx.de> At the end of many or all HTML files in re.search there is no newline. Instead, after the '>' of '', there is this (quotes excluded): ' ' Is this intentional? What is the purpose of this? Why is there no terminating newline? Rainer From rainer.blome at gmx.de Sat Aug 16 21:42:40 2008 From: rainer.blome at gmx.de (Rainer Blome) Date: Sat, 16 Aug 2008 23:42:40 +0200 Subject: [search-ui] Warning regarding "background-color:none;" Message-ID: <48A749D0.5070201@gmx.de> For the URL http://re.search.swlabs.org/recent.html Seamonkey complains in the script console: Warning: Expected color but found 'none'. Error in parsing value for property 'background-color'. Declaration dropped. Source File: http://192.168.1..../.../cool/kt_files/global_changes.css Line: 86 Line 86 currently looks like this: background-color:none; "none" is not a valid value for "background-color". "none" is a valid value for "background-image". "transparent" is a valid value for "background-color". Since I don't really know CSS, and don't know the intent of this line, I don't know how to fix this. The line is part of ".rating-count img {..}". If no background color is desired, that line should be left out, shouldn't it? Rainer PS: For the same page, there is another warning, but that was easy to fix. The cause is a missing left brace: Warning: Expected identifier for pseudo-class or pseudo-element but found '11px'. Ruleset ignored due to bad selector. Source File: http://re.search.swlabs.org/kt_files/style.css Line: 182 .blurb font-size:11px; cursor:pointer; color:#49499a; } From rainer.blome at gmx.de Sat Aug 16 23:57:36 2008 From: rainer.blome at gmx.de (Rainer Blome) Date: Sun, 17 Aug 2008 01:57:36 +0200 Subject: [search-ui] svn lock on cool/search.html? Message-ID: <48A76970.2000600@gmx.de> As discussed in chat, I added