The only thing I really have to say about AO3's latest contretemps is: I toldyou so~
When a platform is so fucked up that the only way to make it run is to take out the navigation, then it's time to think real hard about who's been driving development.
The tag filtering sidebar is currently turned off, and I can only imagine what the tags table must look like if cutting that out of the loop has more or less fixed the site load issues. You can still navigate among fandoms, but any index you just have to haul through story by story, now.
*dry* At least the 1000 work limit has been removed.
*snerks* Yeah, I think that would be diagnostic all right.
This is why Tumblr doesn't actually /index/ tags under a certain threshhold or index back in time indefinitely. When you have that many trash-tags, serving as meta rather than as actual searchables, you really have to make some choices.
Yeah, user-formable tags is deeeeefinitely sketchy space for indexing. It was something I actually considered in hypothetical for my Drupal-based cosplay site.
I saw that announcement and head-bonked against the nearest wall-surface a few times. Oy. Idiots. Their system never really worked all that well in the first place for trying to find anything, and to find out that was the major problem in their server load (which yes, you had predicted well in advance)...
I'm also particularly ticked off by the fact that they have *OVER 17,000* users waiting for an invitation and they're only giving out 100 a day! And they took away our own right to give out invites. WTF? Seriously, you're going to make people wait over SIX MONTHS to get accounts, and in most cases, that's just so they can *read* stories? ???? That's some seriously stupid planning and way to alienate a whole ton of people.
Six months? When the site has been up and running for *years*? At this point, the "beta" should be out of it's name altogether! But they keep hiding behind the "beta" saying "work in progress" when it's really such a slapped together effort.
Not to mention the other thing I found and screamed about earlier (at least they responded promptly to my ticket and admitted guilt in the matter).
Argh. Again, love the concept. Love that they've archived stories. Nice to have... but what a really horrible way of implementing it.
Because I kept meaning to get a AO3 account and putting it off, I figured I'd take this opportunity to join the never ending queue, and this was the little automated response - You've been added to our queue! Yay! We estimate that you'll receive an invitation around 2012-12-17.
Oh yeah, let us cheer that we're going to make you wait half a year. There is not enough facepalm in the world. I mean, at this point, they could at least take ten minutes to change the reply to "sorry" instead of "Yay!".
... That's a really stupid little response. Honestly, folks, take two minutes and change the message. Yay! Six months! {rolls eyes}
If you want one sooner, I do actually have an invite left that they haven't taken away. I don't know what will happen after I send it - they might actually be blocking the ones we send personally. But worth a shot, if you want.
Honestly, it's fine, this encourages me to try to convince them to change their invite system, because I'd have been okay with a month or two wait considering how long the queue is, but 6 months is absolutely ridiculous and something there needs to change.
^_^ "Here, see, I was *almost* a potential consumer, but your shitty policies have turned me off. Change, damnit!"
When I was poking around at the invites, I noticed they now have over 19,000 people on the list. Six months just grew to seven. Pretty soon, they'll be at a year!
Yep! I told them /that/ too, not too very long ago. And it's not like accounts are what adds load (as has been demonstrated). It's not even like /uploads/ are what adds most of the load. *shakes head* Such a mess.
They just proved beyond a doubt that it has nothing to do with the number of people, with that lovely little graph of server load vs their story finders... So let people in! Who knows, with more people, they might even get more volunteers who can help them figure their way out of this mess.
In my industry, we talk a lot about "user experience" but... removing the navigation? That's an experience, certainly, but not one I'd ever wish on any users. What are they smoking over at AO3, anyway?
The sad part? That sidebar full of tags, with which to filter a fandom full of stories, appears to have been responsible for almost two thirds of the server load.
So now they're back to olde timey FFN sort of navigation, where you just keep paging back through the whole list of a fandom twenty at a time. Because the new and improved way to find stories was just that totally broken.
The tag filters pull data from the 1000 works themselves, which is what they were assuming the page load/cache issues were stemming from. (I am not in any way a dba, so have NO idea how much else might be part of the issue and wouldn't even attempt to venture a guess.)
Tags and tag wrangling themselves are unaffected. Meaning - if you click on a tag - everything connected to that tag will pull up.
Advanced search, however, doesn't work that way - if you search for 'Coulson' in the tag search, you'll only get works tagged with 'Coulson', but if you then click on 'Coulson', you'll get works tagged as all the variations attached to 'Coulson'. (Unless something's changed dramatically that I've not been told about.)
So it's not the 'tags' themselves causing the issue, it's the db walking through each of the works that pulls up in the 1000 results to populate the filter w/ the top-level/canonical tags. Which, if it's one or two people, not a problem, but when it becomes waaaaaay more than that - yeah, problem.
How to fix that? *shrug* Again - I am not a dba of any fashion. I wouldn't even know where to start.
I do know that there are plans to update the filtering system. I don't know what they are or what the ETA is - though it's possibly going to get bumped up considerably given the recent evidence, but that's not my domain.
As far as invites - I'm with ya'll there - just send 'em out and let people in. Though, I don't think an invite is needed just to read on the archive - only if one wants to comment or bookmark w/in the archive.
*hollowly* An individual query. For every tag. On every story. In the whole set.
*puts her head down and just cries*
Oh my /god/, who thought that was a good idea?! How did the fact that the query had to be capped not suggest instantly that the basic concept was flawed, and it would only get worse?
How they fix it is to only put the canonical character and relationship tags associated with a given fandom source in that sidebar. Those to be retrieved with a /single/ query to a new table which uses the fandom source id as the primary key to associate canonicals with the source(s). Faceted filtering by freeform needs to wait until they have got a competent project manager and admin, and probably be a totally separate page in the Tags area.
Oh my unholy fuck. I think I need to go get my smelling salts. No wonder it was such a hideous drain.
This would explain why the sidebar had turned into an unusable mess in fandoms that had had crossovers or drabble dump posts dumped into them. It's not actually that they'd associated all the crossover tags in the database, which always struck me as flat out legit insane and something that really made no sense even for their tag wrangling lunacy, it's that they're walking through all the 1,000 works and pulling the tags from them, so -- you get all the characters from the fandoms that are in no way actually associated the fandom in question.
Exactly. As soon as someone said "starting from the story tags" it all made sense. In a really dreadful way. It's exactly because none of the tags are associated with anything but individual stories.
That graphic. That graphic is what is getting me. They put out there in visual form for all to see how flamingly goddamn incompetent they are.
And they probably think that's a good thing because it's totally a sign that they're Improving Communication and Working On The Problems.
(No, what would be improving communication and working on the problems would be recognising that you have profound and fundamental design problems and at minimum hire (yes, HIRE, as in pay them an appropriate professional fee) a consultant to help you go through the code and identify the problems. And then telling us you're doing that, and why. Or hey, mark keeps offering a few hours FOR FREE. Take THAT up and let us know you're doing so.)
I did not know that Mark had offered some of his time... I may chat w/ our Volunteers chair and see if she can coordinate with him on some stuff. I can't guarantee anything at all, but I can point her in his direction. Thanx for pointing that out.
I cannot recall specific details but, yes, mark has indicated several times he'd be happy to consult, particularly since OTWarchive is the only other relatively known majority-female open source project out there and, well, solidarity.
To date, he's been brushed off as far as I am aware, and since he's not personally invested in the platform sufficient to fight to be able to assist them, it hasn't gone anywhere. Perhaps the word in the right ears, though! I hope something comes of it. :)
Oh jesus the fuck christ, the implications really just hit.
Even if they are caching the works listing, they can't be caching the sidebar if they're doing it like that. There's -- I think I need some damn smelling salts myself, now.
I am wistfully wishing they had been associating all the tags that ever get posted in association with a fandom in the database.
Because that would be an absolutely bizarre act of lunacy, sure, but that would be a case of what else is new hey guys the sky is blue and AO3 handles tags bizarrely.'
But... no.
I am all the fuck out of can. Also, even. I am straight up out of even. I CANNOT EVEN, BRANCH. I ACTUALLY CANNOT EVEN.
I did some testing a few weeks ago, and as of then, they were not caching works listings, or if there was any caching it was a matter of seconds. Every single time I clicked on a page in my "1000 works" list, it was doing a new db search. And every single time, it was regenerating the tags list for the filter. Every time I changed sorting. Every page. Every time I used the filter. Crushing the db again and again. It was jawdropping.
That's my take on how it was explained to me. Looking at the graph, I'm almost certain that's exactly what happens - which explains the load.
*does not code for anything, does a LOT of guessing based on logic*
Personally, I love the *idea* of the filters, but I do agree they are clunky as hell. (I'm assuming others felt the same, hence the redesign that is in progress.)
I. Wow. I'd wondered when the load times suddenly dropped -- despite the last update I'd seen reading, in essence, 'we can't do anything for a few weeks, sorry' -- what they'd figured out.
That, though? I did not expect that, and am a little amazed that I didn't notice. (I guess I mostly use in-fic tags, not sidebar tags? Huh.)
Mostly, though? I just want to give a hearty shout of "I SECOND THE MOTION!" to your past statement of tag wrangling on AO3 should be a function that creators and viewers have immediate, front-end access to, because. YES. Yes, yes, and YES IMMEDIATELY.
Seriously, I can *promise* that, when I tag a fic, I am well aware of terms that my tags should overlap with! (And ones they shouldn't.) For that matter, I sometimes find fics when reading whose tags need adjustment/linkage in ways I recognise! Do you even know how many times I have wished for tag-wrangling powers, or even a flippin' SUGGESTION BOX FOR TAGWRANGLING?! I can't even. That. ARGH.
I was really hoping there could be a version on the Upload form, rather than the generic feedback; but I guess that's just too advanced for an archive in beta. It'd be really awesome if it could warn you when a tag was unique, or have a little "this should be linked to" box, something like that. Remembering what I wanted linked and where so that I can go to the Feedback page afterward is kind of difficult, because I have a memory like Swiss cheese.
Still, at least there's one way to suggest that "x should be linked to y" -- thank you!
(That was the incident that opened my eyes to how fucked up tag-wrangling is on the AO3, how badly handled it is. It's the reason I listened so much during the election to other people's problems with tag-wrangling in its current form and came to hate the whole damn thing.)
Have used that twice to explain that while X and Y had the same name, they were from totally different fandoms and ask if there was any way to fix that.
The first time, I eventually got a nice reply saying "Oops, we'll sort that" and the names became Name(Fandom One) and Name(Fandom Two).
The second time I got a snotty reply about how they could only go on what names were there, and nothing could be done about it. When I grumbled about this on Twitter, I got a long lecture from a total stranger (friend of a friend) about how hard it was to be a tag-wrangler and how I had no right to complain because I wasn't even a user. (I'm not registered. And given it takes six months to sign up...)
*snorts* Yeah. The concept and current system behind tags is such a bizarre mashup of "we have to let everyone do anything they want" and "we have to be in complete control". Certainly non-author wrangling should have to go through review/approval, otherwise... oh god, the potential for vandalism. *winces at the thought* But let the /author/ have an "approve it/trash it" button! I mean, if she doesn't want to use it, fine, leave it for the wranglers (poor sods), but give her one! This could go so much faster and smoother with a tiny bit of forethought and automation. And someone in charge who's less of a flaming control freak.
This is so stupid and annoying that I don't even know what to say about it. So, instead I'll comment on the fact that apparently an alphabetical list of writers/accounts was not at aaaaaall helpful for finding ppl I was wondering if they had an account there (Oh, I recognise that name! I wonder what stories they've uploaded? *click*), but a list of 10 random accounts is. Yeah.
*laughs* Okay, that's funny. Yeah, there are so many accounts that going alphabetically takes forever, and the search was pretty useless last time I used it. Way too limited, couldn't even handle "same name with or without spaces". I think, to start with, it was even case sensitive. But the random button likes you! *giggling*
This is probably a stupid question, because I know jack all about site programming, and also the answer is probably "bullshit politics", but...Why can't they just do the tags like delicious? That's super useful and user-friendly, and they don't seem to have the site problems AO3's current model is giving them.
Not that I can give an authoritative answer, because the real answer is buried in black-box meetings. But I think there's a horrible confluence of "the interface must be simple and uncluttered!" and "it will only upset people if we show them how we scramble around and hack at things trying to make this all work" and "/obviously/ we can do it /better/ (despite being totally inexperienced and having an apparent aversion to asking or looking up how to do shit)".
So most of the tag tree is utterly invisible to the users and the wranglers have to bust their asses trying to make users' (forcibly ignorant) choices work sensibly, and never mind the right and left hands, the freaking thumb doesn't know what the first finger is doing. A lot of the essential infrastructure that delicious employed is probably in place--it's just invisible and being manually shoe-horned by an artificial extra layer of labor because no one can see what they're doing and all the users are driving blind. One reason that crowdsourced databases work is that people can see what other people have done, and a general consensus emerges about how we label things. AO3 deliberately shut that off.
Ideally, this particular issue is vast and bad enough that it may prompt some re-evaluation of that rat's nest, but personally I doubt it.
So basically they decided to take a perfectly fuctional and useful system, made it more complicated, cut out the parts that made it work in the first place, replaced them with volunteers who may or may not have the information (or interest) to make useful connections in tagging and are generally treated like shit regardless, and then deal with their clusterfuck by removing basic functionality and also massively limiting people's access even though number of users isn't even the major factor?
Yeah, that about covers it, really. Oh, and didn't take the help of, like, top-ten-in-world type of experts, when it was offered. Oh, and don't forget the part about "wrote code so incredibly bad that generating a sidebar formed two thirds of the site load".
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
branchandroot
no subject
Date: 2012-06-13 07:15 pm (UTC)no subject
Date: 2012-06-13 07:42 pm (UTC)*dry* At least the 1000 work limit has been removed.
no subject
Date: 2012-06-13 08:03 pm (UTC)Wow. Okay. That's painful.
no subject
Date: 2012-06-14 09:14 am (UTC):O
At least the 1000 work limit has been removed
Well, I did really want that -- but not like this. XD
no subject
Date: 2012-06-13 10:13 pm (UTC)no subject
Date: 2012-06-13 10:17 pm (UTC)This is why Tumblr doesn't actually /index/ tags under a certain threshhold or index back in time indefinitely. When you have that many trash-tags, serving as meta rather than as actual searchables, you really have to make some choices.
no subject
Date: 2012-06-13 10:26 pm (UTC)no subject
Date: 2012-06-13 10:51 pm (UTC)Speaks for itself, really.
no subject
Date: 2012-06-13 10:53 pm (UTC)no subject
Date: 2012-06-13 11:27 pm (UTC)Is that table not even partitioned? Are they just calling the whole bloody thing for every sidebar and parsing after that? Good lord.
no subject
Date: 2012-06-14 12:29 am (UTC)no subject
Date: 2012-06-14 01:29 am (UTC)no subject
Date: 2012-06-14 07:30 pm (UTC)no subject
Date: 2012-06-14 12:10 am (UTC)I'm also particularly ticked off by the fact that they have *OVER 17,000* users waiting for an invitation and they're only giving out 100 a day! And they took away our own right to give out invites. WTF? Seriously, you're going to make people wait over SIX MONTHS to get accounts, and in most cases, that's just so they can *read* stories? ???? That's some seriously stupid planning and way to alienate a whole ton of people.
Six months? When the site has been up and running for *years*? At this point, the "beta" should be out of it's name altogether! But they keep hiding behind the "beta" saying "work in progress" when it's really such a slapped together effort.
Not to mention the other thing I found and screamed about earlier (at least they responded promptly to my ticket and admitted guilt in the matter).
Argh. Again, love the concept. Love that they've archived stories. Nice to have... but what a really horrible way of implementing it.
yeaaaaaaaaaaah....
Date: 2012-06-14 12:35 am (UTC)That? That's just absurd.
Re: yeaaaaaaaaaaah....
Date: 2012-06-14 01:27 am (UTC)Oh yeah, let us cheer that we're going to make you wait half a year. There is not enough facepalm in the world. I mean, at this point, they could at least take ten minutes to change the reply to "sorry" instead of "Yay!".
Re: yeaaaaaaaaaaah....
Date: 2012-06-14 07:54 pm (UTC)If you want one sooner, I do actually have an invite left that they haven't taken away. I don't know what will happen after I send it - they might actually be blocking the ones we send personally. But worth a shot, if you want.
Re: yeaaaaaaaaaaah....
Date: 2012-06-15 01:21 am (UTC)Re: yeaaaaaaaaaaah....
Date: 2012-06-15 03:18 am (UTC)When I was poking around at the invites, I noticed they now have over 19,000 people on the list. Six months just grew to seven. Pretty soon, they'll be at a year!
no subject
Date: 2012-06-14 01:25 am (UTC)no subject
Date: 2012-06-14 07:56 pm (UTC)no subject
Date: 2012-06-14 02:15 am (UTC)no subject
Date: 2012-06-14 03:15 am (UTC)So now they're back to olde timey FFN sort of navigation, where you just keep paging back through the whole list of a fandom twenty at a time. Because the new and improved way to find stories was just that totally broken.
no subject
Date: 2012-06-14 02:19 am (UTC)Yea, the tagging system's always been rather bloated but I actually USE those tags to try to filter out some stories I'm looking for...
no subject
Date: 2012-06-14 03:19 am (UTC)no subject
Date: 2012-06-14 04:24 am (UTC)Tags and tag wrangling themselves are unaffected. Meaning - if you click on a tag - everything connected to that tag will pull up.
Advanced search, however, doesn't work that way - if you search for 'Coulson' in the tag search, you'll only get works tagged with 'Coulson', but if you then click on 'Coulson', you'll get works tagged as all the variations attached to 'Coulson'. (Unless something's changed dramatically that I've not been told about.)
So it's not the 'tags' themselves causing the issue, it's the db walking through each of the works that pulls up in the 1000 results to populate the filter w/ the top-level/canonical tags. Which, if it's one or two people, not a problem, but when it becomes waaaaaay more than that - yeah, problem.
How to fix that? *shrug* Again - I am not a dba of any fashion. I wouldn't even know where to start.
I do know that there are plans to update the filtering system. I don't know what they are or what the ETA is - though it's possibly going to get bumped up considerably given the recent evidence, but that's not my domain.
As far as invites - I'm with ya'll there - just send 'em out and let people in. Though, I don't think an invite is needed just to read on the archive - only if one wants to comment or bookmark w/in the archive.
no subject
Date: 2012-06-14 04:04 pm (UTC)*puts her head down and just cries*
Oh my /god/, who thought that was a good idea?! How did the fact that the query had to be capped not suggest instantly that the basic concept was flawed, and it would only get worse?
How they fix it is to only put the canonical character and relationship tags associated with a given fandom source in that sidebar. Those to be retrieved with a /single/ query to a new table which uses the fandom source id as the primary key to associate canonicals with the source(s). Faceted filtering by freeform needs to wait until they have got a competent project manager and admin, and probably be a totally separate page in the Tags area.
Oh my unholy fuck. I think I need to go get my smelling salts. No wonder it was such a hideous drain.
no subject
Date: 2012-06-14 06:10 pm (UTC)no subject
Date: 2012-06-14 06:14 pm (UTC)no subject
Date: 2012-06-14 06:24 pm (UTC)That graphic. That graphic is what is getting me. They put out there in visual form for all to see how flamingly goddamn incompetent they are.
And they probably think that's a good thing because it's totally a sign that they're Improving Communication and Working On The Problems.
(No, what would be improving communication and working on the problems would be recognising that you have profound and fundamental design problems and at minimum hire (yes, HIRE, as in pay them an appropriate professional fee) a consultant to help you go through the code and identify the problems. And then telling us you're doing that, and why. Or hey,
no subject
Date: 2012-06-15 03:06 am (UTC)no subject
Date: 2012-06-15 09:53 am (UTC)To date, he's been brushed off as far as I am aware, and since he's not personally invested in the platform sufficient to fight to be able to assist them, it hasn't gone anywhere. Perhaps the word in the right ears, though! I hope something comes of it. :)
no subject
Date: 2012-06-14 06:39 pm (UTC)Even if they are caching the works listing, they can't be caching the sidebar if they're doing it like that. There's -- I think I need some damn smelling salts myself, now.
No wonder it's two-thirds of their load, god.
no subject
Date: 2012-06-14 07:41 pm (UTC)no subject
Date: 2012-06-14 08:29 pm (UTC)Because that would be an absolutely bizarre act of lunacy, sure, but that would be a case of what else is new hey guys the sky is blue and AO3 handles tags bizarrely.'
But... no.
I am all the fuck out of can. Also, even. I am straight up out of even. I CANNOT EVEN, BRANCH. I ACTUALLY CANNOT EVEN.
no subject
Date: 2012-06-17 10:37 pm (UTC)no subject
Date: 2012-06-15 03:13 am (UTC)*does not code for anything, does a LOT of guessing based on logic*
Personally, I love the *idea* of the filters, but I do agree they are clunky as hell. (I'm assuming others felt the same, hence the redesign that is in progress.)
no subject
Date: 2012-06-14 07:58 pm (UTC)no subject
Date: 2012-06-14 06:54 am (UTC)That, though? I did not expect that, and am a little amazed that I didn't notice. (I guess I mostly use in-fic tags, not sidebar tags? Huh.)
Mostly, though? I just want to give a hearty shout of "I SECOND THE MOTION!" to your past statement of tag wrangling on AO3 should be a function that creators and viewers have immediate, front-end access to, because. YES. Yes, yes, and YES IMMEDIATELY.
Seriously, I can *promise* that, when I tag a fic, I am well aware of terms that my tags should overlap with! (And ones they shouldn't.) For that matter, I sometimes find fics when reading whose tags need adjustment/linkage in ways I recognise! Do you even know how many times I have wished for tag-wrangling powers, or even a flippin' SUGGESTION BOX FOR TAGWRANGLING?! I can't even. That. ARGH.
no subject
Date: 2012-06-14 03:30 pm (UTC)Support and Feedback, choose "Tags" in the first dropdown box.
no subject
Date: 2012-06-14 05:24 pm (UTC)I was really hoping there could be a version on the Upload form, rather than the generic feedback; but I guess that's just too advanced for an archive in beta. It'd be really awesome if it could warn you when a tag was unique, or have a little "this should be linked to" box, something like that. Remembering what I wanted linked and where so that I can go to the Feedback page afterward is kind of difficult, because I have a memory like Swiss cheese.
Still, at least there's one way to suggest that "x should be linked to y" -- thank you!
no subject
Date: 2012-06-14 06:29 pm (UTC)Of course, they may choose to ignore your knowledge of the fandom and the actual real use of various tags, and tell you what's best by giving you a tag no one knows about and will never use and not fixing the actual problem you pointed out.
(That was the incident that opened my eyes to how fucked up tag-wrangling is on the AO3, how badly handled it is. It's the reason I listened so much during the election to other people's problems with tag-wrangling in its current form and came to hate the whole damn thing.)
(Dropping in from link-hopping)
Date: 2012-06-20 04:13 pm (UTC)The first time, I eventually got a nice reply saying "Oops, we'll sort that" and the names became Name(Fandom One) and Name(Fandom Two).
The second time I got a snotty reply about how they could only go on what names were there, and nothing could be done about it. When I grumbled about this on Twitter, I got a long lecture from a total stranger (friend of a friend) about how hard it was to be a tag-wrangler and how I had no right to complain because I wasn't even a user. (I'm not registered. And given it takes six months to sign up...)
So, the option exists, but...
no subject
Date: 2012-06-14 04:10 pm (UTC)no subject
Date: 2012-06-14 09:04 am (UTC)no subject
Date: 2012-06-14 04:14 pm (UTC)no subject
Date: 2012-06-14 10:37 pm (UTC)THIS IS WHAT HAPPENS WHEN YOU ARE CARGO CULT PROGRAMMERS.
no subject
Date: 2012-06-14 10:59 pm (UTC)no subject
Date: 2012-06-15 06:23 pm (UTC)no subject
Date: 2012-06-15 06:49 pm (UTC)So most of the tag tree is utterly invisible to the users and the wranglers have to bust their asses trying to make users' (forcibly ignorant) choices work sensibly, and never mind the right and left hands, the freaking thumb doesn't know what the first finger is doing. A lot of the essential infrastructure that delicious employed is probably in place--it's just invisible and being manually shoe-horned by an artificial extra layer of labor because no one can see what they're doing and all the users are driving blind. One reason that crowdsourced databases work is that people can see what other people have done, and a general consensus emerges about how we label things. AO3 deliberately shut that off.
Ideally, this particular issue is vast and bad enough that it may prompt some re-evaluation of that rat's nest, but personally I doubt it.
no subject
Date: 2012-06-15 07:19 pm (UTC)...Seriously?
no subject
Date: 2012-06-15 07:27 pm (UTC)It is a clusterfuck of truly epic proportions.
no subject
Date: 2012-06-15 07:51 pm (UTC)