pony request: auto link obvious urls in posts September 20, 2013 7:41 PM   Subscribe

I'm not a huge voice on this site but I'm a regular visitor. I often read, and occasionally comment from my phone. It'd be nice if the URLs added in comments were auto linked, it would make it easier for both posters and readers. I understand that their can be ambiguity in URLs but maybe we could make some simplifying assumptions, such as anything matching http://(optional.)(whatever).(whatever) gets linked. I'll keep visiting either way.
posted by askmehow to Feature Requests at 7:41 PM (55 comments total) 2 users marked this as a favorite

I feel like we've talked about this before, but a quick search isn't pulling anything up. I believe the reason we haven't is that the parsing is more complicated than it looks, but pb will be able to address it more specifically.
posted by restless_nomad (staff) at 7:46 PM on September 20, 2013


Agreed that parsing is tough especially if you shoot for 100% accuracy, in this case the 80/20 rule should be more than sufficient.
posted by askmehow at 7:52 PM on September 20, 2013


pb on this previously, in 2012
posted by jacalata at 7:58 PM on September 20, 2013 [1 favorite]


in this case the 80/20 rule should be more than sufficient.

I hear you, but having a new feature that only works 80% of the time isn't really something we'd do. I know the site seems old and creaky in a lot of ways but everything we have on the site functions predictably, and we're sort of fastidious about having it be that way, pb in particular.
posted by jessamyn (staff) at 8:08 PM on September 20, 2013 [7 favorites]


The other issue, which may be minor, is in cases of someone wanting to reference a site without linking them (so as not to give them better page ranking or whatever).
posted by Night_owl at 8:15 PM on September 20, 2013 [1 favorite]


I experimented with this a while back while building an RSS reader for MetaFilter threads. pb's qualms from jaculata's link weren't really an issue, especially when you take into account that people already flag broken HTML to be fixed. It's relatively easy to tell the difference between URLs as elements of links and bare URLs as text in any language with an HTML parser. For any URL that's below a certain threshold of confidence, you just leave it unconverted.

On the other hand, no one was deliberately trying to break my detection scheme, which would probably happen almost immediately if this were ever actually implemented. But for comments written in good faith, I can't think of an occasion when it hasn't worked for me as intended, and I've been using it to read MetaFilter threads for over a year.

Seems to me more like an editorial decision. To Night_owl's point, do you respect the intention of the writer or do you prioritize the experience of the reader? Since I'm primarily a reader, I chose to throw the intentions of the former group under the bus.
posted by Jeff Howard at 8:26 PM on September 20, 2013 [1 favorite]


http://is.gd/JUVbM2
posted by carsonb at 8:35 PM on September 20, 2013 [3 favorites]


You horrible man.
posted by Going To Maine at 8:39 PM on September 20, 2013


Too easy. Though this is as good an opportunity as any to post URLs in bad faith if you want to test edge cases.
posted by Jeff Howard at 8:43 PM on September 20, 2013


I'd be concerned that this might make spamming/SEO bullshit easier (maybe with obfuscated urls?), but I'm not exactly sure about the details of how such a scam might work.
posted by double block and bleed at 8:49 PM on September 20, 2013 [1 favorite]


You shouldn't be posting bare links anyway. You should be posting a hrefs.
posted by Justinian at 10:10 PM on September 20, 2013 [3 favorites]


Just in case you ever wondered what A HREF stood for...

(and how does one escape markup in a comment?)
posted by Annika Cicada at 10:41 PM on September 20, 2013 [8 favorites]


I am sort of against this pony, as I'm a crotchety geek against Do What I Mean on general principles. Also this would make it easier for people to just dump URLs in their text instead of making the link a proper part of the sentence, as God and Sir Tim intended.
posted by Dr Dracator at 11:04 PM on September 20, 2013 [11 favorites]


Won't somebody please think of the wombats ?
posted by y2karl at 11:20 PM on September 20, 2013


Would it be helpful (possible?) to add a little more function to the link maker we already have for those who are confused by markup? Example here.
posted by maggieb at 11:49 PM on September 20, 2013 [1 favorite]


You shouldn't be posting bare links anyway. You should be posting a hrefs.

Exactly. 'Don't plop raw URLs, you lazy sod' should be added to the guidelines.
posted by jack_mo at 11:53 PM on September 20, 2013 [7 favorites]


This sounds like a job for a greasemonkey script, if you ask me. Of course, greasemonkey doesn't work on phones so maybe not. Maybe it's a job for an optional feature that is designed in a conservative way so as to only auto-link links that the URL parser feels confident about, while ignoring links that look difficult to properly parse. I would probably use such a feature on my phone, though not on my laptop.
posted by Scientist at 12:14 AM on September 21, 2013


I've never had a problem just typing in the html for a link on my smart phone and I hate typing on that thing. The most tedious part is copy/pasting the url and that part doesn't change even with some automatic thing going on. The a href syntax isn't exactly difficult.
posted by shelleycat at 2:09 AM on September 21, 2013


That, and the link making button is so easy to use. I don't understand why anyone would want this, but then I don't understand using a phone for internet, so there's that.
posted by Too-Ticky at 4:17 AM on September 21, 2013


I can already double-click on the link to highlight it, then right click and select "Open Link" in Chrome, Safari, and Firefox. (Not sure about mobile; maybe someone can tell us.) I don't find raw links to be an obstacle to reading. But I do find them ugly and would rather they're not encouraged (as well as bolded in a bright color if they're auto-linked).
posted by zennie at 4:31 AM on September 21, 2013 [1 favorite]


(and how does one escape markup in a comment?)

You do it the same way that you would if you were writing a plain HTML document, since that's just what you're doing: writing HTML. In HTML you use &lt; if you want a literal < and not the beginning of a tag, and similarly for &amp; for & and so on. It's not strictly necessary to use &gt; for > but you can if you want. Wikipedia has a good list of HTML named entities.

Also, for some really old reason MetaFilter does not allow numeric entities because I guess it makes it easier to filter dangerous markup without having that avenue of attack available. This is in general the reason why many other systems use a completely different markup scheme (like BBcode) so that anything even remotely resembling HTML gets escaped, as it's fantastically complicated to filter HTML so that it's safe.
posted by Rhomboid at 6:22 AM on September 21, 2013


Bare links are good for those of us who prefer to read metafilter from pages our secretaries have printed out for us.
posted by dawkins_7 at 6:54 AM on September 21, 2013 [12 favorites]


I am quite fond of Metafilter's interface - I want to push back against the idea that it's old and creaky, in case that sentiment results down the line in a conversion to some web 7.0 monstrosity. I like my metafilter, in all it's gritty, hard-edged, hoverlabel-free, basic-search, text-based glory.

Also, get off my lawn.
posted by Salvor Hardin at 7:43 AM on September 21, 2013 [17 favorites]


So, could we characterize metafilter as an artisanal website, full of aging hipsters honing the fine craft of early-aughts web-surfing?
posted by Annika Cicada at 7:52 AM on September 21, 2013 [6 favorites]


Also this would make it easier for people to just dump URLs in their text instead of making the link a proper part of the sentence

Which has always been my main objection, yeah. As it is we deal with bare urls when we do see them by repairing them ourselves, as in not just turning the bare url into a self-similar bare link but by tossing it into an href with some link text to discourage the look of visual clutter and the idea that HEY JUST PASTE BARE LINKS GO CRAZY is an encouraged part of the aesthetic here.

It takes slightly more effort to make a proper link. That's okay. It takes slightly more effort to do a lot of the things that I'm glad people on this site are willing to make that slight effort to do; learning how to use the link tool or write up some markup requires a little learning but not a lot and you're a better web citizen for having done that learning, so hey.
posted by cortex (staff) at 7:58 AM on September 21, 2013 [12 favorites]


I, who know next to nothing at all about HTML, find the link tool extremely easy to use (even from my phone!) and am eternally grateful for it. I just wish all sites had a nifty link button. But I do agree that it's pretty annoying when I have to copy and paste a URL to go see what someone is talking about (especially from my phone!).
posted by Weeping_angel at 8:43 AM on September 21, 2013


I actually have a question about the link tool: why does it default to having the "http://" already completed? Do people use it more for typing in URLs manually rather than pasting them? I've only ever pasted in there and unless I'm missing something, I have to delete it before being able to paste the URL.
posted by griphus at 9:03 AM on September 21, 2013 [6 favorites]


Oh wait I see it selects the text when you do it on an actual computer so you can paste right over it.

Would there be a way to have it default to not putting the "http://" in there if you're in mobile view? It doesn't do that on mobile.
posted by griphus at 9:12 AM on September 21, 2013 [2 favorites]


IIRC the form automatically removes the double http:// upon form submission?
posted by Annika Cicada at 9:18 AM on September 21, 2013 [1 favorite]


To be fair to metafilter, the comment form of the website is so fantastically simple and easy to use. I at one time was an advocate of getting all "web 2.0" with the linking in comment form, but have since recanted my position.
posted by Annika Cicada at 9:23 AM on September 21, 2013 [1 favorite]


IIRC the form automatically removes the double http:// upon form submission?

Yes, if your URL starts with http://http:// the link button code will change it to one http://.
posted by pb (staff) at 9:25 AM on September 21, 2013 [7 favorites]


How about a filter to remove profanity too? That'd be clbuttic.
posted by iotic at 10:08 AM on September 21, 2013 [2 favorites]


As it is we deal with bare urls when we do see them by repairing them ourselves, as in not just turning the bare url into a self-similar bare link but by tossing it into an href with some link text

I'm surprised to hear that. Other than altering Amazon referral codes (and maybe stripping out stormfront links) I was under the impression that you treated the content as sacrosanct from a moderator perspective. I'm a fan of creative editorializing if it encourages people to be more thoughtful about their lack of href text.

From what I can tell, this pony isn't about teaching people HTML or changing the behavior of the contributors in any way (to the extent that's even possible). It's just about making the links that slip through the cracks easier to use for platforms like the OP's phone where it's presumably not so easy to copy and paste bare URLs. If you can't change people's behavior then how do you solve that problem?

cortex's approach addresses it in an unanticipated way. A human-powered solution is always going to be more accurate as long as people regularly flag bare URLs.
posted by Jeff Howard at 10:26 AM on September 21, 2013


I was under the impression that you treated the content as sacrosanct from a moderator perspective.

Mostly true. If we linkify something, something like this:
It's in this book here. http://www.example.com
usually gets turned into one of these:
It's in this book here.

It's in this book here. Link
posted by LobsterMitten (staff) at 10:44 AM on September 21, 2013


Heh. I just linkify the bare link. So if you see inconsistencies, it's because I'm the laziest one.
posted by restless_nomad (staff) at 10:53 AM on September 21, 2013 [4 favorites]


I have a burning hatred for programs that think they are "smarter" than me. If I've put an unURLed link into a comment it's because I don't want that address linked. So I'm voting against this pony being set free.
posted by Mitheral at 10:56 AM on September 21, 2013 [5 favorites]


Yeah, what LM said. I take great pains not to create even subtle editorial implications when I linkify a bare URL; there's usually a natural candidate for link text (a key verb or noun phrase in the preceding sentence), and in cases where there is no such thing I'll use text like "link" or in the case of multiple links "1, 2, 3".

It's just about making the links that slip through the cracks easier to use for platforms like the OP's phone where it's presumably not so easy to copy and paste bare URLs. If you can't change people's behavior then how do you solve that problem?

But making it easier in that specific case by encouraging through automation the unwanted behavior is its own kind of problem. I know there's sometimes value in the notion of paving the cowpaths, but sometimes you don't want the cowpaths to become the official thoroughfare.
posted by cortex (staff) at 10:58 AM on September 21, 2013 [2 favorites]


Interesting. In the previous thread on this topic, jessamyn mentioned that you fix these on average about once per day (c. April 2012). Maybe the solution here isn't to have the system automatically fix bare URLs, but to have the system automatically flag bare URLs to your attention rather than depending on the community to flag them for you. In that case, false positives are much less of a problem.

I totally get that you don't want to reinforce norms for posting bare URLs. My counter is that not fixing them, that is, allowing bare URLs to persist as written, doesn't seem to be much of a deterrent. But you've introduced a better alternative. Fixing the bare URLs properly (rather than automatically) seems like the way to go as long as you've got the bandwidth for that type of thing.
posted by Jeff Howard at 11:12 AM on September 21, 2013


I've never found it all that hard to use the 'link' button from my phone or Nexus 7. The hardest part was always deleting the extra http:// which I now find out isn't necessary.
posted by octothorpe at 11:16 AM on September 21, 2013


In the previous thread on this topic, jessamyn mentioned that you fix these on average about once per day

It feels like a lot less now, though maybe that's just because the edit window means we're not doing any other typo fixes.
posted by jessamyn (staff) at 11:23 AM on September 21, 2013


I think maybe both the reduction in overall typo requests and actually a bit of people stopping and fixing up their own bare links, something I've caught a couple times and smiled at. I think being able to edit has maybe led to a few folks posting a bare url, coming back to the thread, being reminded that it doesn't auto-link, and making the effort to toss an href around it properly. For which, kudos due; way to be, folks.
posted by cortex (staff) at 11:43 AM on September 21, 2013 [4 favorites]


The hardest part was always deleting the extra http:// which I now find out isn't necessary.

Look Marge, I've just tripled my productivity!
posted by arcticseal at 12:33 PM on September 21, 2013


The other issue, which may be minor, is in cases of someone wanting to reference a site without linking them (so as not to give them better page ranking or whatever).

In those very rare instances, there are easy workarounds, e.g.:

www . foxnews . com
posted by John Cohen at 1:15 PM on September 21, 2013 [3 favorites]


> How about a filter to remove profanity too? That'd be clbuttic.

To take the comment seriously for a moment (and illuminate the problem for non-programmers)... The only pragmatic approach is to hire a subscription service to prefilter/flag text streams from the vox pop before your own site processes them for presentation. It's easier to subscribe to a service that keeps a full-time staff working 24/7 to maintain and update the string parsers necessary to suppress objectionable words and phrases in multiple languages than it is to hire the people necessary to attempt this madness yourself.

I did not realize such a service existed until a client insisted that their homepage run a real-time stream from Twitter displaying any tweets bearing a hashtag from a list they believed relevant to their business, and could not be dissuaded from it even after we presented examples that some of them were also used by kids-these-days when tweeting about partying hard.
posted by at by at 4:38 PM on September 21, 2013 [2 favorites]


Eh, depends on the setup. Certainly most basic forum software has a reasonably effective profanity filter built in, if your only goal is to adhere to some arbitrary standards. (E.g. ESRB ratings, which apply to game's official websites too.) They can't do a damned thing about people yelling at each other, but they can make sure they at least spell things creatively.
posted by restless_nomad (staff) at 4:48 PM on September 21, 2013


I'd be concerned that this might make spamming/SEO bullshit easier (maybe with obfuscated urls?), but I'm not exactly sure about the details of how such a scam might work.

I don't see how it enables a scam exactly either, but I think it does exacerbate a common pitfall for unsophisticated web users — a link whose text is an URL really looks like it must be a link to that URL, but it doesn't have to be at all (e.g., http://www.metafilter.com). But the more links you see where the text is an URL and that URL really is where the link goes, the more you come to expect that, and (if you don't know much about HTML) maybe you think that's just how the web works. And as cortex said about the cowpaths, if we automatically linkify URLs, there might be a lot more links here whose text is the URL being linked to, which nourishes this confusion and maybe makes relatively unsophisticated web users more vulnerable to spammers and phishers making links (here or elsewhere) that look like they go to trustedsite.example.com but actually go to hiveofvillainy.example.com.

Also, more bare URLs would be a gigantic change to the aesthetics of the site.
posted by stebulus at 5:39 PM on September 21, 2013 [3 favorites]


Am I the only one who uses Metafilter to generate links for other places? I just pop open a New Post page and make my text all nice and a href'd, then copy and paste elsewhere. The idea of doing bare links here is absurd.
posted by anotherpanacea at 7:53 PM on September 21, 2013 [2 favorites]


Night Owl, for that there's Do Not Link.
posted by IndigoRain at 8:50 PM on September 21, 2013


I, too, dislike link-shortening services.
posted by box at 6:30 AM on September 22, 2013


anotherpanacea, I love when you post workarounds. It's so eponysterical. <3
posted by maryr at 8:02 AM on September 22, 2013 [1 favorite]


Metafilter: gritty, hard-edged, text-based glory
posted by John Cohen at 10:14 AM on September 22, 2013


Am I the only one who uses Metafilter to generate links for other places? I just pop open a New Post page and make my text all nice and a href'd, then copy and paste elsewhere. The idea of doing bare links here is absurd.

Me too, except I use a comment box. I'm always terrified that I'll hit Post by mistake, but so far have managed not to do that (Command-X helps). Very useful when commenting at other blogs that allow a certain amount of HTML but provide no tools for doing it. I'm sure there are other sites/tools out there for doing this, but since I almost always have a MeFi tab open, this is the easiest way.
posted by beagle at 10:31 AM on September 22, 2013


old and creaky in a lot of ways but... functions predictably

We are all Metafilter.
posted by urbanwhaleshark at 3:17 PM on September 22, 2013 [1 favorite]


I have a burning hatred for programs that think they are "smarter" than me. If I've put an unURLed link into a comment it's because I don't want that address linked.

I don't deny that it happens, but it seems to me (based on my subjective, nonscientific observation) that unlinkified URLs are about 98% due to laziness/ignorance/accident and about 2% deliberate choice not to link the URL. Admittedly, I hang out on the Green more than the Blue so it may be somewhat higher on the Blue.
posted by DevilsAdvocate at 7:01 AM on September 24, 2013


unlinkified URLs are about 98% due to laziness/ignorance/accident

My back-of-envelope calculations is that they're like 50% "I'm posting from my phone and don't want to hassle" and then the other half is split into don't know how and don't care.
posted by jessamyn (staff) at 7:12 AM on September 24, 2013


« Older wfrgms has passed away   |   Fantasy NHL 2013 is On Newer »

You are not logged in, either login or create an account to post comments