bubersson / humla Goto Github PK

Create wiki documentation about architecture, installation and stuff.

We agreed on removing MongoDB from Humla's design because it requires running a database server and Humla should be as easy to install as possible. So I did some research and found some possible replacements:

• Alfred (http://pgte.github.com/alfred/) - probably #1 for me, looks pretty capable
• tabler (https://github.com/aarong/tabler/) - offers support for several kinds of backends ( now probably AmazonDB and JSON files, but MySQL is on the way), so it could fit to more users
• perstore (https://github.com/kriszyp/perstore) - similar to tabler

There is also node-dirty (https://github.com/felixge/node-dirty), but according to author it's effective only for <= 1mil records...

I think having a document-based DB could be both useful and easy to install since you need to install only the node module using npm. What do you think?

We should create new (or update existing Lecture 1 of Web Services and Middleware) sample presentation so it will contain new features (usage of gbooks extension, loading elements one-by-one etc). This could be done everytime someone pushes new extension/features to repository

We should merge appliaction design of files newcourse.html, editlecture.html... with mainpage design (index.html) so it would come together nicely.
Also note that after we got login the "Editor" tab won't be displayed.

Form validation is not working properly, multiple elements with same ID etc. Will be fixed over the weekend

We need to have overlaying layer over whole screen on slides, where menu, comments, editor and stuff could be displayed.
It should depend on currently selected View.

(I did a bit of work there)

Just a reminder for me to implement index refreshing when presentation is edited. Will be done as soon as I finish editor view

Directory for index files should be out of public directory. The files should be "hidden" and accessible via API. I'll redesign in this week. The question is where to move it. I would suggest something like /cache/index/... in root of the humla project (aka on the same level as lib, public, handlers). Petr, what do you think?

First we need to make Humla reasonably stupid-proof and then it could be pushed to Node Package Manager registry.
This may cause new users and that may lead to some bugreports and stuff.

Have a look at /handlers/editor.js, about line 425. If I make request in method refreshIndexFile() to node itself, it fails with "Error: ENOTFOUND, Domain name not found". I tried following URL (valid):
http://127.0.0.1:1338/api/MI-MDW/lecture1/index?refresh=true
127.0.0.1:1338/api/MI-MDW/lecture1/index?refresh=true

So for now as a fallback it's handled by client. After Humla will be deployed to production environment, I suggest try to check this issue.

Create a HTML page with documentation of all REST routes. On page should be something like the picture on this site: http://blog.nodejitsu.com/create-nodejs-web-services-in-one-line .

I think link to api.html could be accessible through link at about.html (only) in new design.

What do you think?

In sources we use both "request"&"response" and "req"&"res" ... I vote for the second option.