|
at least there are now RFC-formatted documents floating around and not a vague google doc describing a fraction of the protocol like earlier
|
#
?
Nov 13, 2018 15:24
|
|
|
|
| # ? Apr 25, 2024 21:21 |
|
|
4 hours to pgdump 300gb lol
|
|
#
?
Nov 13, 2018 15:35
|
|
|
redleader posted:did that, uh, help? at all? We needed it in order to make users' avatars show up reasonably quickly, without extra database hits. The thing is, the rest of the avatar settings (shirt type, color, hair style, color) were stored in a different cache, separate from the users' genders. This led to an interesting bug where a user clicked stuff too fast and got a bearded lady avatar, because some code path wasn't properly invalidating the cache. It was necessary at the time, to keep the underpowered database server from melting, but no, it did not "help," seeing as we kept getting the cache invalidation and refresh logic wrong. Eventually, we switched to user-uploaded images as avatars and stopped reading from the caches, which hung around for probably over a year before we finally pruned the dead code. DELETE CASCADE posted:Was this cache primarily used to fulfill search requests for Females to harass Thankfully, nobody tried to add Gender to the search index. A bunch of other Member Profile fields made it in there, I think, and had to be reindexed frequently enough that we had to put it in a separate index so the rest of the site's index could ever keep anything in its caches. And then the users still complained that "search sucks." cinci zoo sniper posted:tfw u nat but it still sucks lol
|
|
#
?
Nov 13, 2018 16:39
|
|
|
CPColin posted:And then the users still complained that "search sucks." nobody is ever happy with search
|
|
#
?
Nov 13, 2018 16:58
|
|
|
Zaxxon posted:nobody is ever happy with search "why is this unstructured data not providing me with automatic insights and organization?"
|
|
#
?
Nov 13, 2018 17:00
|
|
|
Zaxxon posted:nobody is ever happy with search One user's complaint was specifically that their content was not returned first when searching for "PHP".
|
|
#
?
Nov 13, 2018 17:28
|
|
|
I think we can all agree that the worst search is the Windows 10 search. Why in the gently caress does Windows 10 bring up results in all languages? WHAT THE gently caress?
|
|
#
?
Nov 13, 2018 17:52
|
|
|
Sagacity posted:"why is this unstructured data not providing me with automatic insights and organization?" also can you sort by best match?
|
|
#
?
Nov 13, 2018 19:01
|
|
|
CPColin posted:One user's complaint was specifically that their content was not returned first when searching for "PHP". once again, could you please sort by best match? (this is my favorite search feature request that I have gotten multiple times)
|
|
#
?
Nov 13, 2018 19:02
|
|
|
Zaxxon posted:once again, could you please sort by best match? "All these results just say 'PHP' in them! How is that more relevant than the content I wrote, which also says 'PHP' in it???"
|
|
#
?
Nov 13, 2018 19:05
|
|
|
the empty freetext search box is the original sin of computer user interface design. such a simple and seemingly innocent thing, and yet it is a constant reminder of the monumental hubris of techbros. the foundation of human knowledge is not the nerd virtue of being able to remember endless bits of pointless trivia, but rather systematic classification. organizing and cataloging information is a skill as old as writing itself, and here we are supposed to ignore all advances in the field since the dawn of time and instead blindly hope we remember the incantation necessary to coax some idiot scoring algorithm into coughing up a somewhat related result. it is fundamentally anti-intellectual. death to freetext search, all hail archival science.
|
|
#
?
Nov 13, 2018 19:15
|
|
|
google was a mistake
|
|
#
?
Nov 13, 2018 19:18
|
|
|
TheFluff posted:the empty freetext search box is the original sin of computer user interface design. such a simple and seemingly innocent thing, and yet it is a constant reminder of the monumental hubris of techbros. the foundation of human knowledge is not the nerd virtue of being able to remember endless bits of pointless trivia, but rather systematic classification. organizing and cataloging information is a skill as old as writing itself, and here we are supposed to ignore all advances in the field since the dawn of time and instead blindly hope we remember the incantation necessary to coax some idiot scoring algorithm into coughing up a somewhat related result. it is fundamentally anti-intellectual. death to freetext search, all hail archival science. ok vannevar
|
|
#
?
Nov 13, 2018 19:26
|
|
|
Captain Foo posted:ok vannevar seriously though if anyone hasn't read As We May Think yet you should, it is good
|
|
#
?
Nov 13, 2018 19:43
|
|
|
the search box is bad because it hampers association when a good search tool should encourage and exploit it is what vannevar bush would have said, and he would have been right
|
|
#
?
Nov 13, 2018 19:46
|
|
|
TheFluff posted:the empty freetext search box is the original sin of computer user interface design. such a simple and seemingly innocent thing, and yet it is a constant reminder of the monumental hubris of techbros. the foundation of human knowledge is not the nerd virtue of being able to remember endless bits of pointless trivia, but rather systematic classification. organizing and cataloging information is a skill as old as writing itself, and here we are supposed to ignore all advances in the field since the dawn of time and instead blindly hope we remember the incantation necessary to coax some idiot scoring algorithm into coughing up a somewhat related result. it is fundamentally anti-intellectual. death to freetext search, all hail archival science. as opposed to remembering the special incantation needed to remember where something is classified ontology alignment is hell
|
|
#
?
Nov 13, 2018 20:14
|
|
|
TheFluff posted:Bush Was Right tbh vannevar bush had very good ideas, i just wanted to post that at u
|
|
#
?
Nov 13, 2018 20:20
|
|
|
ComradeCosmobot posted:as opposed to remembering the special incantation needed to remember where something is classified yospos is not the place to post nuanced takes but if i were in a charitable mood i guess i might concede that the problem is really more like having a freetext search as the only significant tool you can approach your index with
|
|
#
?
Nov 13, 2018 20:25
|
|
|
sounds like a problem we can solve with machine learning
|
|
#
?
Nov 13, 2018 20:46
|
|
|
redleader posted:sounds like a problem we can solve with machine learning This happened at Experts Exchange, too, to drive the "other content related to this content" stuff. Imagine trying to train an algorithm when nobody could agree on what made two bits of content count as being related to each other. Now imagine that applying the algorithm burned through all our CPU credits!
|
|
#
?
Nov 13, 2018 20:54
|
|
|
TheFluff posted:yospos is not the place to post nuanced takes but if i were in a charitable mood i guess i might concede that the problem is really more like having a freetext search as the only significant tool you can approach your index with stuff like slack now has free text field and the ability to set a bunch of parameters. it's not like i understand the core problem here well but that seems like a nice middle ground
|
|
#
?
Nov 13, 2018 21:08
|
|
|
gonadic io posted:stuff like slack now has free text field and the ability to set a bunch of parameters. it's not like i understand the core problem here well but that seems like a nice middle ground well, not really. if a search box lets you search for e.g. records in a given date range then that's very good but it's not the kind of free text box i'm harping about. free text search is a very blunt tool, or perhaps more aptly put, a way too sharp tool. it sorta works if you know exactly what you're looking for and can come up with sufficiently distinct keywords. it's almost useless for exploring a topic, finding related records, or cross referencing in general. again, it doesn't encourage association. one of the most heinous examples i can think of is spotify. their free text search is almost completely useless, of course, but what really pisses me off is that they have shittons of metadata that they do their best to stop you from exploring, because they want to show you their own "top recommendations" or whatever the gently caress the big record companies want you to listen to. perhaps it is more symptom than disease, though. it's a very common programmer thing to not really think about record classification or metadata and just shovel user input into unstructured text fields - if you're lucky you get both creation and update timestamps (but almost never previous versions), and then a free text search is really the only way to approach your dataset (you can't even speak of an index). a humble tagging system is a very powerful search tool, but only if the metadata is carefully curated and maintained - ask any librarian. really, computer nerds ought to talk more to librarians and archivists.
|
|
#
?
Nov 13, 2018 21:58
|
|
|
TheFluff posted:well, not really. if a search box lets you search for e.g. records in a given date range then that's very good but it's not the kind of free text box i'm harping about. free text search is a very blunt tool, or perhaps more aptly put, a way too sharp tool. it sorta works if you know exactly what you're looking for and can come up with sufficiently distinct keywords. it's almost useless for exploring a topic, finding related records, or cross referencing in general. again, it doesn't encourage association. one of the most heinous examples i can think of is spotify. their free text search is almost completely useless, of course, but what really pisses me off is that they have shittons of metadata that they do their best to stop you from exploring, because they want to show you their own "top recommendations" or whatever the gently caress the big record companies want you to listen to. oh yeah i see what you mean. the slack search i mentioned only even shows you stuff if you can specify it exactly, i.e. " just shovel user input into unstructured text fields" (except some of them are numeric but whatever)
|
|
#
?
Nov 13, 2018 22:08
|
|
|
ultimately, the only correct search results are the ones the user actually wants to see. but it's impossible for a user to directly beam their actual intentions into a system, and even if it was, it's impossible for the system to understand whatever vague unbounded poo poo is going through the user's head sometimes i want to search for a specific citation from a specific source on a specific date. sometimes I want to search for "I vaguely remember hearing about an article on this subject five years ago, I'm going to mash somewhat-related half-remembered search terms in until I see something that looks familiar" the ideal search system depends on what kind of information you store, what kind of user you expect to be using the system, and so on. it's a UX decision first and foremost, not really a code decision...if you live in a magical utopia where everyone does what makes sense and managers always stop to think things through
|
|
#
?
Nov 13, 2018 22:11
|
|
|
Main Paineframe posted:the ideal search system depends on what kind of information you store, what kind of user you expect to be using the system, and so on. it's a UX decision first and foremost, not really a code decision...if you live in a magical utopia where everyone does what makes sense and managers always stop to think things through Designers often see search as a way around having to think about the user experience at all, and they then condition users to expect search to be magic. I've literally seen dudes in talks with customers saying poo poo like "oh well just make the search better, then you don't have to do any tagging"
|
|
#
?
Nov 13, 2018 23:39
|
|
|
That also happened at EE. Somebody got a boner for tags, but didn't really tie any functionality other than search to them. Content was already categorized by "topic," which is what determined which "top experts" leaderboard you got on when you answered questions. There was no explanation of how tags and topics were different, no reason given why users could create arbitrary tags, but not arbitrary topics, etc. Everybody started complaining about stuff being mis-tagged and people finally realized that when you've got somebody with a tech problem stumbling onto Experts Exchange to ask how to use Google, you were practically guaranteed to have some lovely tags added! So they got rid of tags and finally allowed users to create their own topics. (Well, sort of. The moderator staff has to do it. At least the whole system somebody proposed of having new topics be "provisional" until enough content came along to promote them to "real" topics got scrapped after about forty hours of meetings about it.) Moral: user-generated content is terrible
|
|
#
?
Nov 13, 2018 23:53
|
|
|
Main Paineframe posted:ultimately, the only correct search results are the ones the user actually wants to see. but it's impossible for a user to directly beam their actual intentions into a system, and even if it was, it's impossible for the system to understand whatever vague unbounded poo poo is going through the user's head in other words... ComradeCosmobot posted:ontology alignment is hell
|
|
#
?
Nov 14, 2018 00:53
|
|
|
CPColin posted:Moral: users are terrible
|
|
#
?
Nov 14, 2018 00:58
|
|
|
CPColin posted:Moral: user-generated content is terrible present company excepted ofc
|
|
#
?
Nov 14, 2018 03:55
|
|
|
pokeyman posted:present company excepted ofc speak for yourself
|
|
#
?
Nov 14, 2018 04:57
|
|
|
TheFluff posted:seriously though if anyone hasn't read As We May Think yet you should, it is good drat
|
|
#
?
Nov 14, 2018 06:21
|
|
|
redleader posted:google was a mistake our intranet uses some "google search appliance" bought at vast expense years ago and it is completely loving useless. like completely random ordering of results, irrelevant suggestions etc etc. I mean our intranet is trash anyway but still, lol
|
|
#
?
Nov 14, 2018 10:23
|
|
|
TheFluff posted:the search box is bad because it hampers association when a good search tool should encourage and exploit it I mean often the path is just google -> wikipedia which is kind of what you want? But it sucks if whatever you're interested in isn't nerdy enough to have a good wiki concept cloud.
|
|
#
?
Nov 14, 2018 11:19
|
|
|
godsdamn no wonder snack overflow ate experts exchange’s lunch
|
|
#
?
Nov 14, 2018 12:15
|
|
|
Powerful Two-Hander posted:our intranet uses some "google search appliance" bought at vast expense years ago and it is completely loving useless. like completely random ordering of results, irrelevant suggestions etc etc. I mean our intranet is trash anyway but still, lol this thing is deprecated, because google doesn’t like maintaining things and also because it loving blows, apparently, I haven’t yet heard about anyone who actually likes the thing
|
|
#
?
Nov 14, 2018 15:25
|
|
|
eschaton posted:godsdamn no wonder snack overflow ate experts exchange’s lunch well, that and snack overflow didn't put a blur filter over the "answers"
|
|
#
?
Nov 14, 2018 15:29
|
|
|
akadajet posted:well, that and snack overflow didn't put a blur filter over the "answers" And didn't ignore its developers when they warned Marketing that doing aggressive SEO to get at the top of every Google results page was actually harming our reputation because everybody was getting so mad at slamming into the paywall and us "gaming the system" to appear so high. Then Google released the "Panda" update and started factoring in how fast your site was, how correct the HTML was, how much people hated your guts, etc., and that was the end of EE's reign at the top of the search results.
|
|
#
?
Nov 14, 2018 16:22
|
|
|
Arcsech posted:this thing is deprecated, because google doesn’t like maintaining things i was talking to a former colleague the other week who said "yeah we were working on a deployment on aws for months then management [of a multinational multi billion dollar org] announced that in future they would only consider using google cloud so we lost the contract ", i just said 'lmao good luck to them with that when Google abandons all support after 18 months" idk why you'd use anything google "provides" for enterprise poo poo
|
|
#
?
Nov 14, 2018 18:48
|
|
|
CPColin posted:That also happened at EE. Somebody got a boner for tags, but didn't really tie any functionality other than search to them. Content was already categorized by "topic," which is what determined which "top experts" leaderboard you got on when you answered questions. There was no explanation of how tags and topics were different, no reason given why users could create arbitrary tags, but not arbitrary topics, etc. Everybody started complaining about stuff being mis-tagged and people finally realized that when you've got somebody with a tech problem stumbling onto Experts Exchange to ask how to use Google, you were practically guaranteed to have some lovely tags added! Oh god we were gonna do that promoting provisional topics thing, but we just told the sales team just tell us to add some every once in a while. They pretty much never do, and I pushed back hard for a while to try and keep the total number of topics at 420.
|
|
#
?
Nov 14, 2018 19:05
|
|
|
|
| # ? Apr 25, 2024 21:21 |
|
|
sharepoint is a perfect case study in why search boxes are poo poo
|
|
#
?
Nov 14, 2018 19:32
|
|
