elz | Code, code, code

Deploy day at AO3 today! We're upgrading from Ruby 1.8 to 1.9, from Rails 2 to Rails 3, and from Apache and Passenger to Nginx and Unicorn. We're also replacing our authorization and authentication systems, changing our html parsing and sanitizing, and updating/replacing a whole bunch of third-party plugins and gems, as well as updating nearly every file in the codebase to work with the new version of Rails. The site will be down for an extended period of time while all the underlying software is switched over.

So, that's not stressful at all, haha! On the plus side: unicorns!

Basically, the number of users and works and the amount of traffic keeps increasing, and we hope to do some hardware upgrades, but in the meantime, we're doing what we can to improve performance and maintainability. (Plug: if you happen to know a lot about MySQL or memcached and you'd like to help us improve our system, we will love you and throw roses at your feet!) An enormous amount of work has gone into this deploy, and I feel like I did barely any of it - yay, team! And the yuletiders who've been helping out with testing are beyond awesome.

My goal for the next couple of weeks: see how much I can get done in terms of filtering and browsing. The filters are both unwieldy and a big drain on performance, so we've been looking to replace them for ages, and the upgrade should facilitate that. And we've been brainstorming about browsing for eons as well - its time has come! *resolve face* I try to use the archive as much as I can, and those things drive me crazy as well. Maybe more so than other people, even, since I stop to yell at myself every time I get frustrated. :)

Flat | Top-Level Comments Only

You and the rest of the coders are awesome.

Also, I'm now curious. Replace the filters with other filters or replace the filters with another method altogether?

Right now, they're the list of all the tags associated with all the works for the exact query you've made, which includes being adjusted for what works you as an individual user are able to access. And the list changes every time you change the filters. It's kind of a monster query, and we haven't been able to cache it because it's so variable. So I'm thinking some combination of a cached list of the top tags for the user/fandom/tag/collection you're looking at and then text fields with autocomplete or some way to type things in to filter by rarer tags.

And ultimately, I think a key is to make it so that the filters aren't the only way to see what tags are out there, so that it doesn't matter so much if they show you a full and exact list. I want to have an easy way to get to my favorite character in a fandom, or my favorite ship, I want to be able to see what ships there are for a given fandom in the archive, etc. And there's got to be a better way to find out that there's zombie fic for The West Wing than scrolling to the end of the six-foot-long sidebar, you know? :)

(Also, fun fact: Sidra ran the numbers, and people only use the current filters about 3% of the time when browsing works. So they're slowing the whole site down, and they're not even getting that much use! /o\)

How accessible is typing in tags, though? I think something visual is needed, even though I understand the performance issues.

(I use the filters a fair bit, and I'm definitely one of the 'see what's out there' crowd.)

Also, I've seen a fair bit of "Delicious is good filtering done well", and I'd like to say, Delicious is unaccessible to me -- there are too many mouse-driven events. My arms get tired from fibro, and I'm quite the keyboard user.

Amy, if you have mousing issues I recommend you switch to Google Chrome and install Vimium. It's what

panrubius uses.

I'm an Emacs girl -- I use emacs editing of textboxes and have YakShave installed in Chrome. Thank you anyway!

If you have a solution, why can't you use delicious? With Vimium, panrubius can use delicious and he cannot use a mouse at all. What is the problem?

(This is not a combative question, I am genuinely interested for accessibility design reasons.)

Because it's not a perfect solution. It's been awhile, but I remember hating the post new bookmark interface, because it was hard to find, and too lengthy to use (the toolbar, as you can imagine, wasn't a great choice). Sometimes all I wanted was to plug in a link and click submit. I remember trying to delete a section of bookmarks (which was all of my bookmarks at the time, since I imported them and changed my mind) and having to go through them page by page, selecting all of them, and then deleting them. Which was wasteful, and as I didn't have the keyboard navigation I have now (Chrome for Linux didn't exist then), it was also painful. There was way too much text on the page, and this upset my Asperger's brain. I need visual aids in the way of graphics, and too much text leads to mental overload on my part.

(I could go back and point out more things that were unaccessible to me, but I don't know how much detail you want. All I can say is that I'm not a fan of delicious' attempts to browse -- the lists, like the AO3, become unwieldy, and the 'bundling' is awkward to use. I remember trying to set it up and getting frustrated at how much clicking was involved.)

/me nodsnodsnods

That's really useful, thank you! <3

Argh, this is really late! I thought I had answered. Anyway, much thanks for the tech-made-easy explanation! It makes sense.

I was curious about the type in filters that made a showing during the yuletide sign-up, but not wanting to badger the archive, didn't tried them. Are they similar to what you envision?

Are you making any changes to the download feature? I was about to file a support request and if there are changes happening there, I'll wait, play a bit, and see if the problem has been resolved prior to bothering you guys?

Here are the things that got fixed for downloads:

-Draft chapters were getting pulled into the download in some formats
-Downloads of single-chapter works no longer have a "Chapter 1" link
-The comment link takes you straight to the comment form now
-The tag section is now included in the download (woot!)
-Downloads should have titles instead of all being "download.epub" if you download them with Stanza
-Titles with non-ascii characters shouldn't cause errors anymore
-Encoding problems in PDF and MOBI downloads should be fixed

I know there are some other formatting/style issues that people have noticed, though, and if there's anything else you've spotted, feedback would be awesome!

The conversation is happening in the

ebooks community here.

It looks like an issue that won't be resolved by the changes you listed. I'm still trying to suss out when the bad behavior happens so that the coder has more data to work with.

Aha! Yeah, it would be awesome if you could send that on through the support form. If there's a way to fix it, we'll definitely give it a try!

I will definitely submit it. :D

Great, thanks for sharing this post. Awesome.

Unicorn!

It's such an amazing piece of work - SO MANY CHANGES!

*pets the unicorn*

I really like and appreciate your article.Thanks Again. Awesome.

Code, code, code

<3

Re: <3

Re: <3

Re: <3

Re: <3

Re: <3

Re: <3

Re: <3

Re: <3

Re: <3

no subject

no subject

no subject

no subject

no subject

cCqOogHFbKoRypO

no subject

no subject

XCGDPYkEQtVQ