The New Heroku (Part 2 of 4): Node.js & New HTTP Capabilities

blog.heroku.com

49 points·joshuacc··6 comments

I deployed HN Trends to Heroku's new Cedar Stack and it was simply awesome!

It stayed on a single dyno the entire time (I tried scaling it to 2 at one point, but quickly realized that my use of an in-memory object that needed to persist between requests was not "shared-nothing") and it handled the rush of traffic with only 1 small hiccup.

I got a little nervous when Tim O'Reilly linked to the site from his Twitter account, but Heroku handled it with ease.

Lack of websockets support is pretty crippling; at least it causes node.js to lose a lot of its value over alternate frameworks... the kind of app that node is perfect for is often also the kind of app that websockets are perfect for.

Nonetheless this is exciting news! Maybe websockets could be added as a Heroku add-on?

Does the lack of websockets in most users browsers also cripple the web? Node's plenty good at HTTP. In fact it's excellent at HTTP. In any case, they are pretty clear about what it will take to get them to add it.

Those are good points. Websockets are just so awesome that it's frustrating that I had to setup a vps to use them in a weekend app :)

What about Socket.io? Heroku points to this announcement and it looks like the sort of thing that could function similar to Websockets.

(I just started with Node so please forgive any glaring ignorance I display.)

Socket.io will work with long polling or flash sockets, so it may end up being the approach people take.

However Websockets is a significantly more efficient, standards-oriented approach. There is a large class of apps that would be really awesome (webkit browsers, websockets + node) that will not be hostable with Heroku. In my personal opinion, these would be really awesome apps.

Reply on news.ycombinator.com