I'm deploying a new (backwards compatible) URL scheme for mefi

February 1, 2007 2:08 PM

I'm deploying a new (backwards compatible) URL scheme for mefi [more inside]
posted by mathowie to MetaFilter-Related at 2:08 PM (136 comments total)

I'm in the process of deploying a new URL format for mefi, and just to head off all the questions and concerns, I wanted to explain the whys and hows of the whole thing here.

What is it?

I'm basically moving from the format of:

http://ask.metafilter.com/mefi/12345

and instead to something like this:

http://ask.metafilter.com/12345/How-to-find-a-lost-dog-in-NYC

The old style URLs will still work, as both are loading the very same file.

Why do this?

There are two reasons. The major one is that I'm constantly finding people on other blogs or even within the MeFi universe say stuff like "I love the answers here" or "This is a great thread" and when you mouse over the links, you don't get any information about it. They could be about anything at all. I noticed that when I hovered over links to blogposts and digg and similar sites, they often stuffed the title into the URL so it was often helpful to figure out what lay ahead on links that were not described.

The second reason is so that questions in Ask MetaFilter will show up in Google more often. Now, that's not an entirely selfish decision and I'm not trying to game google, but when you search for a full question's text on Google, An ask mefi post may show up in the top 50-100 spots but rarely closer than that. I realized last week when a dishwasher died at my house, when I did searches for what dishwasher to buy, I got nothing but pointless product pages for the first 100 results when a search, but when I searched for the term diswasher at Ask MeFi, I got a really good answer. We have 40,000+ questions on every subject under the sun now, so we might as well help others out with all this information.

How's it going to work? What about X?

The structure of links will be whatever.metafilter.com/-id number-/-title of post-, and since the ID is within the URL, no matter how URLs get cut off in email or other devices, the proper post will show up. Old URLs will work forever as well, but references throughout the site will eventually all follow the new format. And while I was going to make sure that full URLs would be less than 80 characters in length to work best with text-wrapping on email apps, I noticed that digg and millions of wordpress and vox blogs all put the full titles in the URL, length be damned. And since cut-off URLs will still work, even if your email app does cut it off (try it) the proper page will come up. All punctuation is removed and spaces are replaced with dashes, making highly readable URLs. It should be a smooth, harmless switch over since the original URLs will continue to work.

Some examples?

Here's a list of 50 questions from today on Ask MeFi. The URL structure looks like the following:

/56046/Click-Click-Boom
/56045/Help-me-move-my-playlist-across-the-file-structure-divide
/56044/How-can-I-sell-data-on-demand
/56043/Why-do-companies-wait-so-long-before-giving-out-W2s
/56042/Can-you-get-New-Mexican-food-in-Phoenix
/56026/How-to-handle-applying-for-work-8-hours-away-from-your-current-residence
/56041/How-to-find-a-job-in-AUSNZ-from-Canada
/56040/Best-brainstorming-practices
/56039/CDDVD-drive-kaput
/56038/Best-lens-for-K10D-camera
/56037/Making-two-columns-work-in-MS-Word-shouldnt-be-this-hard
/56036/Home-Network-Expansion
/56035/Webform-for-item-tracking
/56034/Breakfast-in-Miami
/56033/Give-me-advice-on-settling-a-bodily-injury-claim-for-a-car-accident
/56032/Help-me-find-examples-for-my-exam
/56031/What-works-of-fiction-are-most-firmly-embedded-in-the-American-cultural-consciousness
/56030/Does-anyone-know-the-lyrics-to-Words-by-Sangie-Davis-Lee-Perry
/56029/Setting-up-a-think-tank
/56028/Is-it-fishnappingor-MURDER
/56027/Client-Server-Database
/56025/How-to-cancel-a-gift-subscription-and-still-get-the-cash
/56024/Print-Audit-alternatives
/56023/How-warm-should-my-landlord-keep-my-apartment
/56022/Help-Me-Write-My-Cheese
/56021/MS-Office-Small-Business-Accounting-Help
/56020/Are-teachers-tougher-than-students
/56011/sentinal-pile-should-i-worry
/56019/Overall-costs-of-thrombosis
/56018/Cheapo-ethnic-jewellery-in-the-UK
/56017/Counselortherapist-in-the-Boston-area-MBTAaccessible-who-is-good-at-handling-career-and-life-goal-questions
/56016/What-films-should-I-watch-to-better-appreciate-Kill-Bill
/56015/Please-help-me-find-a-replacement-for-my-HP-Jornada-minilaptop
/56014/Firefox-extension-to-control-HTML-forms
/56013/Mapping-drive-to-MSSQL-table
/56012/Unique-reception-venue-in-Sioux-Falls
/56010/Insurance-PreExisting-Condition-Woes
/56009/A-banana-would-have-made-it-perfect
/56008/Where-can-I-buy-Napalese-jewelry-outside-Nepal
/56007/Should-I-pursue-him

There are joke answers to be sure, but the vast majority are really obvious and helpful. An undescribed link to them reveals what they'll likely be about.
posted by mathowie (staff) at 2:09 PM on February 1, 2007

I like. Makes sense.

From what you've said, it seems that all of the posts in MeFi's history will be revamped with new URLS. Am I understanding this correctly?
posted by defenestration at 2:17 PM on February 1, 2007

I notice the older/newer links don't have the new URL (yet).

It'd be nice to have a way to get the 'long form' url when looking at a page via the short form.
posted by These Premises Are Alarmed at 2:18 PM on February 1, 2007

You are only doing it for electoral purposes, aren't you? When are the next elections around here?


jokes aside, sounds really good.
posted by micayetoca at 2:21 PM on February 1, 2007

And, I don't know if this is new or not, but there are some funny characters showing up at the bottom of Ask index pages (firefox and safari):

√
posted by These Premises Are Alarmed at 2:22 PM on February 1, 2007

Cool.

This breaks the Mefi Navigator script (I'm not bitching, I realise this is my problem). Fixing it may not be entirely trivial, because of the way GreaseMonkey deals with URLs, but it'll begin working again soon.
posted by matthewr at 2:22 PM on February 1, 2007

This is a good idea. Even better if people stop using quirky, cute titles for their questions that don't give you any idea what the question is about. Although there isn't a whole lot of that going on in those 50 questions above.
posted by bob sarabia at 2:23 PM on February 1, 2007

Question: aside from maintaining existing old-style URLs, will new posts also receive an old-style URL along with the new?
posted by cortex at 2:25 PM on February 1, 2007

The only problem I can see for it is that, unlike Blogger or Digg, it allows you to write whatever you want in the URLs. That's great for email programs that cut things off, but it's also great for writing links like this or this. I'm not even sure that's much of a problem, but it's definitely something that lots of people will do.
posted by Partial Law at 2:29 PM on February 1, 2007

Yeah, for plugin people, you just need to tweak your URL grepping, so that instead of looking at the mefi.com/mefi/## part of the URL, you'll just need to grab the ID from the previous "directory" in the structure (so mefi.com/##/foo instead).

The next/back links will be updated to the new ones. Basically, in a few hours, every link to a thread anywhere on mefi (except metatalk for now, with no titles) will have this new structure.

The old structure will work and will work forever (it's just an apache thing loading the same file), but all references will be to the new format.
posted by mathowie (staff) at 2:30 PM on February 1, 2007

Very nice.
posted by OmieWise at 2:34 PM on February 1, 2007

Not to be a smart-ass, but if
http://ask.metafilter.com/12345/How-to-find-a-lost-dog-in-NYC
gives me the same thing as
http://ask.metafilter.com/12345
it's just ignoring everything after the number. Fine.

So will it work if I just stick in random junk there?

http://ask.metafilter.com/12345/What-you-going-to-do-with-all-that-junk-all-that-junk-inside-that-trunk

?

Could be used to redirect improperly and/or construct misleading and only marginally amusing jokes about threads (probably by me).
posted by GuyZero at 2:38 PM on February 1, 2007

It's as though I've entered a fantastical dreamworld where ponies appear before I can even imagine them.
posted by koeselitz at 2:38 PM on February 1, 2007

This is a good idea. Even better if people stop using quirky, cute titles for their questions that don't give you any idea what the question is about.
posted by ThePinkSuperhero at 2:39 PM on February 1, 2007 [5 favorites]

GuyZero - I just tried that with

http://ask.metafilter.com/56046/Click-Click-Boom

and

http://ask.metafilter.com/56046/anything.

They seemed to go to the same place...
posted by muddgirl at 2:41 PM on February 1, 2007

Yeah, I lose a lot of sleep over that problem, too.
posted by Dave Faris at 2:42 PM on February 1, 2007

it's just ignoring everything after the number.

Effectively. However, as mefi itself will be generating a single, content-specific url, and that is what Google will be indexing, random misleading/jokey/inconsistent links deployed by others on now and then don't really affect anything.
posted by cortex at 2:42 PM on February 1, 2007

Oh - I had read Matt's post as boing in the future tense. Well, consider that answered. I don't know if it's harmful, but this question does.

I'm honestly not trying to make trouble, but if doing this causes any sorts of issues, it should be dealt with. It shouldn't be a big deal unless some high-traffic site starts generating bogus MeFi URLs that get indexed by Google.
posted by GuyZero at 2:44 PM on February 1, 2007

yeah, it's basically like how I used to do magical subdomains like mathowiesucksballs.metafilter.com, and yeah you can do that now, but since it has to be a link that gets indexed I think it's more of an outside, hopefully seldom used joke.

Digg and wordpress require the entire title no matter how long, but I figured that'd break too many times in email clients so I just ignore the text when displaying a post. If the joke use of it gets out of hand, I could make sure the title matches up with the post title and spit errors on other stuff, but I hope I don't have to do that.
posted by mathowie (staff) at 2:47 PM on February 1, 2007

Me Likey. Me also likey how GuyZero has managed to find a nefarious use for this particular pony already.
posted by Sk4n at 2:48 PM on February 1, 2007

Even if it were a big deal, it wouldn't be a new deal. Consider the humble anchor notation.
posted by cortex at 2:48 PM on February 1, 2007

cortex: true. But I wonder if Google even indexes anchor text?

Anyway, sounds like the right call vs what Digg, et al do.

Also, I think I broke something trying this link. If you delete the text at the end it gives a JDBC Crash - like this.
posted by GuyZero at 3:01 PM on February 1, 2007

"Error Occurred While Processing Request: Variable FRIENDLYURL is undefined."

Mozilla/5.0 (Macintosh; U; PPC Mac OS X; en) AppleWebKit/418.9.1 (KHTML, like Gecko) Safari/419.3
posted by Blazecock Pileon at 3:04 PM on February 1, 2007

GuyZero, both of those are bad URLs. There's no /mefi/ in the new structure. Also, the old URLs never ended with a slash.
posted by mathowie (staff) at 3:09 PM on February 1, 2007

Blazecock, that error existed for a short period while we were working on the page, it should be gone now.
posted by mathowie (staff) at 3:10 PM on February 1, 2007

Sweet. May I suggest a (permalink) somewhere on the page?
posted by Skorgu at 3:13 PM on February 1, 2007

Oh. Huh. I see how it works for AskMe but it doesn't seem to work for MetaFilter proper yet, perhaps by design.

And I see what you mean about the training slash. It was actually a typo on my part. It's not an issue generating crashes? If not, no problem.

I should get back to filing bugs against my own software now...
posted by GuyZero at 3:14 PM on February 1, 2007

Blazecock, that error existed for a short period while we were working on the page, it should be gone now.

Yep, works now. Sweet.
posted by Blazecock Pileon at 3:16 PM on February 1, 2007

But I wonder if Google even indexes anchor text?

Good point. I have no idea how to google for that, either. Heh.
posted by cortex at 3:21 PM on February 1, 2007

It's not like the text matters, after all, this works:

monkey.metafilter.com/comments.mefi/58261
posted by blue_beetle at 3:22 PM on February 1, 2007

I like this idea. It makes sense, is helpful and has no real downside thanks to the backwards compatability.

I would suggest that something is mentioned on each posting page about writing an informative title, with specific reference to the fact that it ends up in the url. The ask posting page already has a nice little comment about what the title should be but the mefi one doesn't, and I think letting the poster know how that field is going to be used would be good. Based on the fifty ask.me's above it looks like uninformative titles aren't really a big problem anyway, but I don't think it will hurt to be extra clear.
posted by shelleycat at 3:23 PM on February 1, 2007

One final idea: could you rewrite the URL so that when someone links to this, the URL that shows up in the browser is the one with the correct descriptive text? That would semi-correct the issue if people started clicking on silly links.
posted by GuyZero at 3:28 PM on February 1, 2007 [1 favorite]

That's a good idea, GuyZero.
posted by cortex at 3:29 PM on February 1, 2007

Matt, how long will it be until Metatalk follows the same scheme?

It doesn't seem like implementing titles in Metatalk is a prerequisite for switching to the new URL scheme, since they work just fine without the title in the URL. People who use the various Mefi scripts and extensions are inconvenienced while the URL scheme is different across the subdomains.
posted by matthewr at 3:31 PM on February 1, 2007

It seems to me that converting this

http://ask.metafilter.com/mefi/12345

to something like this:

http://ask.metafilter.com/mefi/12345/How-to-find-a-lost-dog-in-NYC

would mean that the meaningful (to software) part of the URL would be the same, which might mean that some software that works with such URLs would continue to work.
posted by Crabby Appleton at 3:33 PM on February 1, 2007

Also, ask.metafilter.com/XXXX gives a 404 error if you don't use a trailing slash.
posted by matthewr at 3:35 PM on February 1, 2007

I'm adding titles to MetaTalk in the next few days.
posted by mathowie (staff) at 3:37 PM on February 1, 2007

This is great. Today, I finally feel I got my five bucks worth.
posted by veggieboy at 3:37 PM on February 1, 2007

Thanks.
posted by matthewr at 3:41 PM on February 1, 2007

I'm adding titles to MetaTalk in the next few days.

Oh glory of glories. That should not please me as much as it does.
posted by cortex at 3:41 PM on February 1, 2007

I'm adding titles to MetaTalk in the next few days.

I look forward to many more of these.
posted by GuyZero at 3:44 PM on February 1, 2007

right, matthewr, because there needs to be a trailing slash and some content afterwards. None of the URLs that people have made up here will be present on the site, so I'm not too worried about fixing bad URLs that the site doesn't generate.
posted by mathowie (staff) at 3:46 PM on February 1, 2007

Fair enough.
posted by matthewr at 3:51 PM on February 1, 2007

[this is good]
posted by dg at 3:56 PM on February 1, 2007

Great! Thanks Matt. Won't help hama7 however.
posted by peacay at 4:08 PM on February 1, 2007

I wonder what the chances of two threads having identical titles are, and URL collision leaving mangled bodies strewn across the highway? Slim to none, I guess.
posted by stavrosthewonderchicken at 4:13 PM on February 1, 2007

So, I ask without an ounce of snark, can someone explain why [this is good]? I mean, it's good for google, I guess, and Matt, and people who can't read the metapages, but why are any of you excited about it?
posted by Dave Faris at 4:18 PM on February 1, 2007

stavrosthewonderchicken: will that be a problem since the numbers in front of that will always be different? I don't know much about the technical stuff like this, but I would think that post "45333/parking-ticket-advice" and "46001/parking-ticket-advice" wouldn't be a problem because the thread numbers are different. But maybe I am not even grasping the problem you are anticipating.
posted by dios at 4:18 PM on February 1, 2007

stavros, the ID is actually the thing that pulls up the thread and those are unique, so collisions won't be possible on titles alone.
posted by mathowie (staff) at 4:19 PM on February 1, 2007

This can be a cool addition as long as people use it to be helpful and not funny.

So, my extreme irony there didn't come across. If you really care, I find the multiple MeTa threads about you ridiculous, dios. I was making a joke.

As I mentioned above, it would be nice to rewrite URLs so that sillyness stopped once the page loaded up. But as Matt pointed out, MeFi itself won't generate these links and it will be a joke that most people don't see, much less use to advance personal grudges. And much like pancakes and ceiling cat, I expect it'll simply pass once people get tired of it.
posted by GuyZero at 4:19 PM on February 1, 2007

So, I ask without an ounce of snark, can someone explain why [this is good]?

It's not a big deal at all, but if you're following links from other blogs to mefi you'll at least know a little bit about the post before you click. Same with links there, when people reference older threads. It's just a bit more metadata in the URL, before you click.
posted by mathowie (staff) at 4:21 PM on February 1, 2007

And if you're not into Digg, et al, it's more like them:

http://www.digg.com/tech_news/Courtney_Love_does_the_math_or_Why_you_needn_t_shed_any_tears_for_the_RIAA
and
http://www.metafilter.com/58258/In-Mission-Control-while-the-loss-of-signal-was-a-cause-for-concern-there-was-no-sign-of-any-serious-problem

No need to guess what that's about, versus the now-outdated:

http://www.metafilter.com/mefi/23251

And it also makes the title field more useful outside of RSS feeds.
posted by GuyZero at 4:25 PM on February 1, 2007

but why are any of you excited about it?

Personally, I'm just excitable. It's a good, if small, change.
posted by cortex at 4:26 PM on February 1, 2007

stavros, the ID is actually the thing that pulls up the thread and those are unique, so collisions won't be possible on titles alone.

Right, gotcha. I missed that part, sorry. It's still first thing in the morning, here.
posted by stavrosthewonderchicken at 4:31 PM on February 1, 2007

Hmmm.

That kind of messing around seems like a Bad Thing (which is (and sorry if I missed a mention of it) that the link http://nastybadthings.metafilter.com/19/a-thread-about-squicking actually works, so it seems as if arbitrary text after http://*.metafilter.com/xxxx/ (where xxxx is the thread number) will redirect to the correct place). Ripe for haha teh funney inside MeFi, but open to abuse outside our little tribe, no?
posted by stavrosthewonderchicken at 4:54 PM on February 1, 2007

I wish youtube would jump on this logical URL wagon.
posted by peacay at 4:58 PM on February 1, 2007

Yes, stav, but it's not hard to mostly fix. Plus, if I interpret Matt correctly, exactly what kind of abuse is it? I don't think it'll make for a googlebomb or anything similar. And hardly worse than this.
posted by GuyZero at 5:02 PM on February 1, 2007

I dunno. People always seem to be able to find ways to commit evil fuckery when the door is left open, though, so I thought I'd raise my hand. It's probably fine.
posted by stavrosthewonderchicken at 5:05 PM on February 1, 2007

Evil fuckery will result in as-yet-unspecified severe penalties, so don't even think about it.
</menacing look>

that ok?
posted by jessamyn (staff) at 5:16 PM on February 1, 2007

:D
posted by Colloquial Collision at 5:21 PM on February 1, 2007

http://www.metafilter.com/username/evil%20fuckery

Invalid user ID

Sweet. I sense a sockpuppet in my future.
posted by quin at 5:30 PM on February 1, 2007

I wish youtube would jump on this logical URL wagon.

Absolutely.
posted by Wolof at 5:36 PM on February 1, 2007

that ok?

Ayup!
posted by stavrosthewonderchicken at 5:38 PM on February 1, 2007

So, my extreme irony there didn't come across.

I wonder how many generations of mankind will have to arise, mature, sicken, and die, passing on their seed and their hard-won cultural knowledge, before denizens of the far future will realize that irony does not come across on computer screens? I guess it would help if we really could push a button on our computer and cockpunch the person on the other end, so that would-be ironists whose words are taken literally have immediate cause to regret their rhetorical faux-pas. Matt, can you get to work on that?
posted by languagehat at 5:51 PM on February 1, 2007 [3 favorites]

mathowie writes "I'm adding titles to MetaTalk in the next few days."
Yes! This will make it a lot easier when multiple tabs are open.
posted by Mitheral at 5:55 PM on February 1, 2007

may i suggest sandwiching some date info in there too, kinda like a blog URL structure?

/56046/Click-Click-Boom

might become


/56046/02012007/Click-Click-Boom

or

/56046/Click-Click-Boom-02012007

I'd love

/2007/02/01/56046/Click-Click-Boom

but that would screw up the thread-number-first convention, and anyway, i know better, I think, than to expect this.

I would note that the lack of URL date info has occasionally given me a bit of greif when employing mousing to trey to know what year a given MeFi link was generated, and moreso now that the thread-counts are so wildy different fom site to site.
posted by mwhybark at 6:15 PM on February 1, 2007

Good call, Matt, thanks.
posted by mediareport at 6:16 PM on February 1, 2007

languagehat: I wonder how many generations of mankind will have to arise, mature, sicken, and die, passing on their seed and their hard-won cultural knowledge, before denizens of the far future will realize that irony does not come across on computer screens?

Or inflection and tone in general? It's an inner voice thing. Because people hear their own voices inside their head giving their words texture and added meaning, they don't realize that their are other interpretations possible for what they wrote. Luckily for me, I speak in a monotonous drone. Irony is one of the hardest registers to master in writing and reading. It's why it gets people into so much trouble. The whole inflection/tone thing is why emoticons where a inevitable byproduct of the internet era.

And as to the topic on hand. The idea got me all giddy and excited, but I can't articulate precisely why. Thanks, Matt!
posted by Kattullus at 6:23 PM on February 1, 2007 [1 favorite]

I just had a mental image of a red-faced angry jessamyn yelling "Evil Fuckery! Evil Fuckery!".

I'm gonna go curl up in a corner now.
posted by pjern at 6:25 PM on February 1, 2007

Y'know, the phrase "Evil Fuckery!" is just calling out to be set to music of some sort . . . do we have anyone here who does that sort of thing?
posted by booksherpa at 6:54 PM on February 1, 2007

Interesting idea, but it seems like the title part of the URL is unnecessary for the actual query.

If there is a question like:

http://ask.metafilter.com/12345/How-to-find-a-lost-dog-in-NYC

will

http://ask.metafilter.com/12345/OHYEAH

Also work?
posted by delmoi at 6:57 PM on February 1, 2007

Can I do the "Metafilter: Evil Fuckery" thing now?

Oh.
posted by Sk4n at 7:03 PM on February 1, 2007

Y'know, the phrase "Evil Fuckery!" is just calling out to be set to music of some sort.

Dammit, I'm already recording an album this month.
posted by cortex at 7:04 PM on February 1, 2007

"I'm adding titles to MetaTalk in the next few days."
I got so disappointed when I realised that this didn't say "I'm adding titties to MetaTalk in the next few days". Because, you know, that may help.
posted by dg at 7:07 PM on February 1, 2007 [1 favorite]

Cortex - To the tune of Mel Brooks' "High Ang......xiety!", please.
posted by Sk4n at 7:09 PM on February 1, 2007

I read this: a dishwasher died at my house, when I did searches for what dishwasher to buy...

And I thought you were buying a PERSON. And I thought, "mathowie is doing preeeeeeetty well for himself if he can buy PEOPLE to wash his dishes"

hmph.
posted by exlotuseater at 7:25 PM on February 1, 2007 [2 favorites]

Please use underscores since dashes between words aren't readable.
posted by nixerman at 7:28 PM on February 1, 2007

languagehat, there's already an irony button on MeFi, it's the little letter I down on the lower right of the comment field. Isn't it? And the B is for bellicose, right? Have I been mistaken this whole time?
posted by sleevener at 7:35 PM on February 1, 2007

delmoi: yes. Please see warning re: evil fuckery, however.
posted by jessamyn (staff) at 7:48 PM on February 1, 2007

Here is a version of Plutor's Metafilter Deleted Threads viewer greasemonkey thing that I've hacked up to work again. It kinda sorta works for Ask also but not really, you get a ton of false positives because of how anon works.
posted by Rhomboid at 7:55 PM on February 1, 2007 [2 favorites]

And I thought you were buying a PERSON.

I thought the exact same thing. And I wondered why there was no mention of calling the police.
posted by A dead Quaker at 7:59 PM on February 1, 2007

I mean, if a dishwasher died at my house I'd be sure to notify the authorities.
posted by A dead Quaker at 8:01 PM on February 1, 2007

Lucky for me, I speak in a monotonous drone.

Secret's out! Kattullus is really a robot!
posted by grapefruitmoon at 8:08 PM on February 1, 2007

Please use underscores since dashes between words aren't readable.

Google does not understand underscores, but does understand dashes.
posted by stavrosthewonderchicken at 8:10 PM on February 1, 2007

Clearly he's been working his servants too hard; seems like they're dropping like flies.
posted by Jimbob at 8:13 PM on February 1, 2007

I think it's a great idea, but I completely agree with GuyZero's warnings: the URL must be rewritten to the one true "correct" one, or it will be abused.

Do the change, it makes the system more useful, but it should be taken as given that

http://ask.metafilter.com/12345/How-to-find-a-lost-dog-in-NYC

will be linked by pranksters as

http://ask.metafilter.com/12345/I-want-to-cook-and-eat-this-dog

Rewriting the URL will mitigate the abuse. People will still do it, but it won't mislead anyone because the correct URL will appear when you bring up the page. If you don't do it from the outset, jokers are going to push the envelope until you feel forced to.
posted by Loudmax at 8:15 PM on February 1, 2007

I guess it would help if we really could push a button on our computer and cockpunch the person on the other end, so that would-be ironists whose words are taken literally have immediate cause to regret their rhetorical faux-pas. Matt, can you get to work on that?

Uh huh, you just want him to crash the server.
posted by Tuwa at 8:16 PM on February 1, 2007

Sounds great! And now, about those new ideas of yours for MeFi Music...?
posted by flapjax at midnite at 8:28 PM on February 1, 2007

the URL must be rewritten to the one true "correct" one, or it will be abused

Argh. This so-called abuse is nonsense. As pointed out above, you can already achieve the same effect if you want to using anchors, and yet nobody has been doing that. You'd have to be a real idiot to look in the URL and see something about eating babies or whatever and take it as true without actually loading the page.

Allow me to give you another example of a huge site that has done it for years this way, without any such care for this so-called "abuse":

http://news.com.com/Why+Apples+board+is+standing+by+Jobs/2100-1047_3-6155451.html

is the same as

http://news.com.com/2100-1047_3-6155451.html

is the same as

http://news.com.com/Steve+Jobs+Eats+babies+ZOMG+Hitler/2100-1047_3-6155451.html

Now, C|NET has been doing their URLs like this for years. And I don't remember anyone complaining about how badly it's abused. And C|NET has probably 10 times the traffic that Metafilter does.
posted by Rhomboid at 8:39 PM on February 1, 2007

I could be wrong, but it doesn't seem that C|NET let random strangers include markup in their contributions to the site, though.
posted by dg at 9:09 PM on February 1, 2007

Great. For people who want the posts' titles on the front pages, it is now trivial to hack up a script to provide them.
posted by MonkeySaltedNuts at 9:17 PM on February 1, 2007

it doesn't seem that C|NET let random strangers include markup in their contributions to the site, though

How is that relevant at all?
posted by Rhomboid at 9:25 PM on February 1, 2007

This is good (for you).
posted by furtive at 9:35 PM on February 1, 2007

World Reels At New Avenue For Evil Fuckery
MeTa-zens Agree—"WHAT LUCK!"

Adding title attributes avec nastygrams to random tags is way more fun anyway.

This is neat mathowie; a real good addition to site functionality. Woot!
posted by carsonb at 10:48 PM on February 1, 2007

For people who want the posts' titles on the front pages, it is now trivial to hack up a script to provide them.

Not if the title is Eats, Shoots & Leaves.
posted by Chuckles at 11:23 PM on February 1, 2007

MetaTalk - Please see warning re: evil fuckery.
posted by Ufez Jones at 11:44 PM on February 1, 2007

A few suggestions:
- lowercase everything
- it's probably a good idea to limit the length (you could perhaps remove things like "how-to-", "why-do-" etc. to help with this)
- do a permanent redirect to the 'authoritative' URL if the title bit isn't precisely correct (including case and trailing slash)

"Argh. This so-called abuse is nonsense."
Attentional to detail isn't nonsense, and it's worth taking a little time to minimise the potential for duplicate/misleading pages.
(I wouldn't use C|NET as an example if I were you, their sites use very poor URLs. Oh, and that site is called News.com, and they own that domain, yet use news.com.com. Crazy.)
posted by malevolent at 3:16 AM on February 2, 2007

The whole OMG URLZ argument is silly – if anything it's just another new outlet for snark being gifted upon us by our lord. It will join:

1. #anchor-snark
2. The title tag
3. Google
4. <RIP> <IMG> <MARQUEE> <BLINK> <INPUT> <SELECT> </RIP>

In the pantheon of snark, dead and living. I didn't know until just now that the form elements were being stripped – oh well, they will be missed, at least until I care enough to bore a tunnel through mathowie's usually poor regexes.
posted by blasdelf at 4:12 AM on February 2, 2007

What's this got to do with Dios?
posted by Brandon Blatcher at 5:18 AM on February 2, 2007

I always hover over URLs before clicking so this will make my life a little bit easier when reading MetaTalk where people always link to things on the blue and the green. Thanks, Matt.

But it would be even sweeter and easier to read if the URLs were all lower case, IMNSHO.
posted by sveskemus at 5:45 AM on February 2, 2007

it doesn't seem that C|NET let random strangers include markup in their contributions to the site, though

How is that relevant at all?

It's relevance seems obvious, am I missing something? Trusted users can be given nice things that untrusted users can't. And matthowie trusts us, for now.


Because, you know, that may help.

*adds titties to projects for world peace and appreciation of ponies day*
posted by and hosted from Uranus at 6:06 AM on February 2, 2007

Actually, blasdelf, you should have thrown <blink> between a pair of <lazarus> &lt/lazarus> tags.
posted by cortex at 6:27 AM on February 2, 2007

Shouldn't you do a 301 redirect from the old-style URLs (i.e. ask.metafilter.com/mefi/NNNNN) to the new sytle URL (i.e. ask.metafilter.com/NNNNN/title-word)? As it currently stands duplicate content is returned at both URLs. If you don't, Google will include both of the duplicate pages in its index and consequently tend to give them lower PageRank than you'd get with a single page.
posted by RichardP at 6:41 AM on February 2, 2007

I've updated Mefi Navigator to cope with the changes. It can be downloaded here.

Userscripts.org is being useless, as usual, and not allowing me to update the script there, or even change the description to point people toward the new version.
posted by matthewr at 7:04 AM on February 2, 2007

I know I'm late to the love fest but I wanted to chime in and say I noticed it this morning and am smitten with the new URLs.
posted by revgeorge at 7:14 AM on February 2, 2007

I thought Blink was still OK.

RichardP writes "Google will include both of the duplicate pages in its index and consequently tend to give them lower PageRank than you'd get with a single page"

And double the number of hits during a site search for the same page. This is already a pain because of archive pages.
posted by Mitheral at 7:19 AM on February 2, 2007

I'm late here, but I just wanted to say this is a good thing. Thanks Matt!
posted by ob at 7:44 AM on February 2, 2007

Shouldn't you do a 301 redirect from the old-style URLs

Yeah, I was thinking the same thing on the streetcar this morning I need a life. There are a lot of potential URLs that will bring up a given page, but when any given page is served it should always be redirected to the "canonical" address for that page.

But I'm not the boss of Matt and I'm sure he'll decide what he wants to do without any more input form me. Just sayin', that's all.
posted by GuyZero at 7:45 AM on February 2, 2007

stav: Google does not understand underscores, but does understand dashes.

I dunno, the link you cite is almost 3 years old, which in internet time is like a decade or something. I would guess google figured out underscores by now. At any rate, digg results (using underscores) keep popping up in my google searches. I personally prefer underscores to dashes for readability.
posted by MetaMonkey at 8:17 AM on February 2, 2007

The rollout still isn't done yet, but by the end of today, there won't be any references to the old URLs on the site, so that shouldn't be a problem.
posted by mathowie (staff) at 8:31 AM on February 2, 2007

The rollout still isn't done yet, but by the end of today, there won't be any references to the old URLs on the site, so that shouldn't be a problem.

But the old URLs will still be in Google's index, so it will still visit them, as well as the new URLs. You definitely want a 301 redirect on the old URLs, otherwise Google will never remove them.
posted by RichardP at 8:53 AM on February 2, 2007

fwiw, on the sites that I manage, we get much better google placement when the topic of the pages is part of the URL.
posted by signal at 9:35 AM on February 2, 2007

links no longer open in new windows.
posted by quonsar at 10:00 AM on February 2, 2007

or new tabs.
posted by quonsar at 10:00 AM on February 2, 2007

Links to threads and other internal MeFi links on the site have always and forever linked with the same window as a target. New windows is only for outside sites in a post or comment.
posted by mathowie (staff) at 10:11 AM on February 2, 2007

ok, let me rephrase that:

links to outside sites in a post or comment no longer open in new windows.
posted by quonsar at 10:24 AM on February 2, 2007

or tabs.
posted by quonsar at 10:24 AM on February 2, 2007

New links should still work. I'll check it out.

And on the redirect thing, I just figured out a fairly easy way to do it without requiring a db hit on every pageview, which I'll implement today.
posted by mathowie (staff) at 10:53 AM on February 2, 2007

It's relevance seems obvious, am I missing something? Trusted users can be given nice things that untrusted users can't. And matthowie trusts us, for now.

I think you must be missing something. We are talking about the ability to link to a page on a site using a URL that can contains any words you want it to, as the words are ignored when loading the page. The actual content on the page is irrelevant here, so I don't see how it matters whether users on the site can post markup or not. I don't require the ability to use an <A> tag on CNET's site in order to send someone a link to a CNET story that contains the words that Steve Jobs eats babies in the URL.

And my point is that people are getting worked up over nothing, because this ability has existed already in the form of #anchors, and it has existed on other sites for a long time and nobody has used it there for pranks (and if they did, nobody noticed.)
posted by Rhomboid at 1:56 PM on February 2, 2007

Ok, I see what you mean now. Sending somebody a spoofy link is different than the spoofy link actually appearing on the site though. More officialicious. But I agree there's some molehill humping going on.
posted by and hosted from Uranus at 2:06 PM on February 2, 2007

Thanks matthewr. I did not know I liked the MeFi navigator so much but I really missed it already.
posted by davar at 2:32 PM on February 2, 2007

I dunno, the link you cite is almost 3 years old

I linked to Phil Ringnalda because he's the FUCKING MAN. As far as I know (and there are fresher discussions of it), Google still doesn't like underscores.

I don't care -- I still use 'em at my sites, because I think the dashes look like ass, and I'm not too worried about SEO bullshit for my meagre output. But one of Matt's stated purposes (or did I imagine it) for doing this was so Google'd see Metafilter better. Dashes (unless there's new information of which I'm unaware) are the better choice for that purpose.
posted by stavrosthewonderchicken at 4:03 AM on February 3, 2007

Matthewr, thank you for realizing it's broken and promising to fix it - I was just e-mailing you to tell you!
posted by IndigoRain at 4:51 PM on February 3, 2007

malevolent: I agree re: the .com.com thing. Only thing I can imagine is they want all of their sites under a unifying domain. Or they think .com.com gives them some kind of leet status because they have, like, the world's most desirable domain or some shit. Either way, it's totally lame. But other than that, I think their URL structure isn't so bad (e.g. no article_id=12345&site=5&cid=2&lang=en nonsense)
posted by heydanno at 7:41 PM on February 3, 2007

As far as I know (and there are fresher discussions of it), Google still doesn't like underscores.

I couldn't resist having a little look, and its seems dashes certainly were preferable a couple years ago, though now the question is debatable, with no particularly persuasive evidence either way.

Gets on my tits that some random quirk of google from years ago has these massive reverbarations for years to come, users be damned. In their position as de facto standard setters, I wish they'd be a little more communicative about these sorts of things. The notion of 4+ years of debate about dashes vs underscores makes my brain ache.
posted by MetaMonkey at 9:44 PM on February 3, 2007

I wish the imdb would do this - i hate having to click on their links to find out what movie/actor the link is referring to.
posted by davey_darling at 9:28 AM on February 4, 2007

I've written a GreaseMonkey script, Mefi Titles 2, that adds the encoded titles to the front pages.

One thing I noticed is that the encoding process loses a lot of punctuation - e.g.

• Web 2.0 ... The Machine is Us/ing Us -> Web-20-The-Machine-is-Using-Us
• Allegations of rape-as-torture at Abu Ghraib surface on YouTube. -> Allegations-of-rapeastorture-at-Abu-Ghraib-surface-on-YouTube

If the reason for encoding is to help Google searches, then those encodings will screw up searches for "Web 2.0" and "rape".

In the script above I also included 2 javascript functions (encodeTitle and decodeTitle) based on encodeURIComponent that fully encode all characters in a title, convert spaces to hyphens, and delimit words with "-"s. They are inverses, i.e.

title == decodeTitle(encodeTitle(title))

For example a title like I ♠ my little "cat-dog"? gets encoded as I--%E2%99%A0--my-little--%22-cat-dog-%22%3F--3q72j18. For this example note in particular that "cat-dog" remains as a unit in the encoding without extra cruft separating the words.

Matt, I haven't been able to figure out the reasons for your encodings. Why is Web 2.0 ... The Machine is Us/ing Us encoded as Web-20-The-Machine-is-Using-Us? encodeURLComponent would seem to indicate that it would be safe to encode it as Web-2.0-...-The-Machine-is-Us-%2F-ing-Us.

Anyway could you update your scheme to make it complete and not create words like rapeastorture? The example functions I provided are based on the standard encodeURIComponent and may serve as a starting point.
posted by MonkeySaltedNuts at 2:26 PM on February 4, 2007

He already said upthread that the algorithm he used was simply to remove all punctuation from the title before converting space to "-", so that's why "rape-as-torture" got converted to "rapeastorture". I assume he was just going for something brainless that wouldn't leave any question as to being valid URLs. I'm sure it could be refined to allow through certain characters but I personally don't see a pressing need. And having %-entities in URLs is fugly to the extreme, please don't advocate that. The point is to convey enough content that it is easy to read at a glance, not to be able to roundtrip the exact characters of the title.
posted by Rhomboid at 4:21 PM on February 4, 2007

I've updated Mefi Navigator to cope with the changes. It can be downloaded here.

Thank you thank you thank you. I was lost without it.
posted by Famous at 4:23 PM on February 4, 2007

This is going to make my previous comment sound snarky, instead of just worthless noise..

I agree about the %-entities, but keeping periods and hyphens in place seems pretty easy and useful (and commas, any others?). I suspect that certain other punctuation would be better replaced with a hyphen than dropped outright, like both slashes.

If we could come up with a nice compact list of rules, it would make it easy for mathowie to implement..
posted by Chuckles at 4:31 PM on February 4, 2007

#Rhomboid: having %-entities in URLs is fugly to the extreme

I don't care if you think "%"s in the URLs is "fugly". Practically nobody looks at them so it doesn't matter if they are "fugly".

If the URI's can encode something meaningful then let them do it.
posted by MonkeySaltedNuts at 6:41 PM on February 4, 2007

Practically nobody looks at them so it doesn't matter if they are "fugly".

The heck are you talking about? The whole reason that Matt made this change was so that when you are presented with a link to a MetaFilter thread you can have an idea what it's about by glancing at it in the statusbar. Not being fugly is the entire point. Nobody cares if some punctuation is lost along the way, but they will care if the URLs turn into uglified %-strewn monstrosities. You're trying to sell the idea that these URLs should look incomprehensible to humans so that the 1 user in 100 that uses an obscure greasemonkey script can see exact titles, while everybody else is out of luck.
posted by Rhomboid at 11:41 PM on February 4, 2007

And yes, I am aware that %-entities are decoded by the browser when displayed in the status bar. But that relies on them being in a link in a browser; it does nothing for the case when somebody emails you a link, IMs you a link, pastes a link in an IRC channel, includes the URL in a comment without using a <A> tag, et cetera. If the point of the exercise is to make URLs that are easily read by humans then let's do that, let's not make them totally cryptic to humans but rely on the browser to decode them for us.
posted by Rhomboid at 11:52 PM on February 4, 2007

I'm with Rhombold on this. The goal was to make an human readable URL, not preserve greasemonkey hacks on the site. MonkeySaltedNuts, you've found two URLs among thousands that do exhibit a problem when I do simple regex punctuation removal, but I can certainly tweak that so dashes are preserved (slashes never will be, because they break the URL scheme).
posted by mathowie (staff) at 8:42 AM on February 8, 2007

I know that you don't fully support lofi.mefi but this seems to have broken the ability to click on (n new) comments in Safari. It always goes to the top of the thread.
posted by tellurian at 6:44 AM on February 16, 2007

Safari doesn't preserve the #anchor part of URLs on a 301 redirect, so all of the existing links to individual comments that use the old URL are now broken. Gah!
posted by chrismear at 4:38 AM on February 19, 2007