I continue to enjoy keeping an eye on these single-page style web apps for very basic content applications. I have a feeling this was pretty quick and easy for the developers to put together.
That said, a few things jump out at me that make me think this approach isn't ready for primetime (or to be more precise, takes a lot more work to make it ready for primetime)
* Individual properties are not reflected in the URL, meaning I couldn't copy and paste a link to someone
* This is using a plain hash instead of a hash bang, which, correct me if I'm wrong, means sub pages are not hit by search engines
* If javascript is disabled and I load this, I get a blank screen.
It's this kind of thing that makes me fearful of pursuing this approach for more serious projects...
> * This is using a plain hash instead of a hash bang,
> which, correct me if I'm wrong, means sub pages are
> not hit by search engines.
Fortunately, you're wrong about hashbangs. They were a temporary hack to allow the GoogleBot to index Ajax-driven pages before the HTML History API existed. Because of New Twitter, many sites have cargo-culted the hashbang without implementing the required "escaped_fragment" functionality on the server-side to make it actually work. It's the web developer equivalent of an SEO rain dance.
Please don't use hashbangs in 2011. Use pushState() if you're a public-facing site that needs to be indexed by search engines.
TIL. Thanks for disabusing me of the notion that this is a good idea. I have seen it used in numerous places which is what led me to assume it was the "right" way.
I assume, then, that the developers of this site could have used pushState() instead of the hash? I mean, I assume that Backbone encourages this style if the developer makes use of it.
Backbone encourages pushState() if you can swing it -- you'll need to implement every client-side route as a server-side page as well ... so it's not possible if you're hosted in an inflexible environment. Backbone falls back to hashes for browsers that don't support the API, and transparently converts URLs between the two.
But regular hashes ("#") are fine for internal single-page-apps that don't need google indexing -- of which the majority of Backbone projects probably are.
It's just the hashbang "#!" that's a particular pet peeve of mine ;)
I don't follow how pushState (being a javascript function) replaces the need for _escaped_fragment contraption for letting crawlers fetch page snapshots in a specific state.
With pushState you can actually push a URL without the hash. It looks like I'm GETting /about, while it's really just / with the content of /about fetched via AJAX.
A HTTP GET /about should of course give the same content as the AJAX version.
Right, but this still doesn't alleviate the need for "prerendering" specific pages for the crawler, which is the most tedious part of the whole thing. In other words, when a crawler requests /about, it should be getting a fully-formed About page with the content already in place, and not a bare container that ajax'es the content on page load (which is how most of pageless sites work).
Yeah - that sucks, but is actually also necessary for optimal performance for users as well. I don't think it's excusable to have two HTTP requests for page content. (and hash bangs are just as bad when loaded directly)
This is mostly right, but there's a very important exception: IE doesn't support pushState. If IE users are an important part of your user base, you should use hashbangs with _escaped_fragment_, but only on IE.
I know I'm showing my age by saying this, but I really don't see the benefit of AJAX in this case. It's basically a frameset powered by javascript, so why not just use a frameset to begin with? It would work in more browsers without issues with back-buttons and take advantage of their built in functionality.
AJAX has its place, but I don't see it here.
Edit: I find it amusing that the documentation for the <frameset> tag is missing :-)
If I am not mistaken the main problem with a frameset would be that the URL in the address bar will never change and therefore you cannot link directly to a page within the frame, i.e. where a user wants to copy the URL and share with a colleague.
* Yes, Individual properties should have been reflected in the url.
Actually if you type in the property in the search box it is creating a url which does the work, but clicking on a property is not making the link.
Hope they'll refine it a bit.
Many single page interface sites do put the parameters in the URL in the fragment. It's kind of a hack, but the link should still work if done correctly. It is still best to use history.js or something that will only resort to fragments when the History API isn't supported.
The real problem is the developer not respecting the href in the anchor and only handling onclick. This prevents you from saving the shortcut easily, opening in a new tab, etc. I like the approach Google+ and others have taken which says "use the href unless left clicked and no modifiers (i.e. ctrl, alt, shift) are pressed.
Since then, with the help of HNer vecter, it's expanded to encompass some more things. The goal right now is to expand this into a more general documentation/code search.
I have instacss.com in my bookmarks bar. Is dochub.io a "better" site to use, or does it not matter which one I use?
Incidentally, I find it very strange that when the site first loads, it's a completely blank page with a Feedback tab on the right. It can sometimes take quite a while for the actual contents to load.
instacss.com or dochub.io will take you to the same place, so it doesn't matter which one you bookmark. We moved the domain just to make the purpose of the site clearer moving forward.
As for the long initial load time, client-side caching is on the top of our list of todos, so stay tuned (or just wait for the HN effect to wear off :)).
It's not just the HN effect. I used instacss.com yesterday to look up the allowed values for -webkit-user-select and was rather surprised at how long it took before the page was usable. That said, the site really is brilliant.
Awesome job! I wanted to do something similar and I'm glad I was beaten to the punch. Keep iterating on the project, the idea itself and the execution has a lot of room for growth and 'betterness'. Kudos.
I created a similar project (CSS docs only) after seeing the original version of this posted here and thinking of some improvements (instacss I think it was called). My version uses the HTML5 history API and runs off of static files (scraped with ruby). If anyone is interested:
Why is it that every time I see a .io domain, I'm not surprised there using the cliche twitter bootstrap. Other than the lack of originality in design, I think it's pretty useful. Although I would like to see some decent HTML5 documentation. It's not unusual for me to search several websites for HTML5 docs. Would be nice to have a 1 stop place for HTML5 and the related languages. Keep up the good work, great project!
There was another "instant docs" site posted to HN a while back that had the same issue this one has: there's no delay between keyup and pushState(), so if I mistype and backspace a few times I've got a dozen entries in my history.
Wow, extremely helpful. Thank you for this tool. Though it did take 2-3 seconds to load (which resulted in me refreshing the page thinking it wasn't working), the documentation is fantastic. I appreciate the contribution!
This is a gem. For someone who just recently learned HTML and CSS (about 6 months ago) I can usually remember just enough to get it wrong. It's handy that you can just start typing and get to the right thing.
I'm hoping this becomes my source of front-end reference goodness! It is quite usable in its current state and has potential for reference of jQuery and other library/API inclusion. Great work!
That said, a few things jump out at me that make me think this approach isn't ready for primetime (or to be more precise, takes a lot more work to make it ready for primetime)
* Individual properties are not reflected in the URL, meaning I couldn't copy and paste a link to someone
* This is using a plain hash instead of a hash bang, which, correct me if I'm wrong, means sub pages are not hit by search engines
* If javascript is disabled and I load this, I get a blank screen.
It's this kind of thing that makes me fearful of pursuing this approach for more serious projects...