server.js should honor Content-Disposition and pass-through header

[garyvidal]

When using rest extensions, the server.js should honor a request with Content-Disposition and forward it as a header back to application

To resolve issue

      //If the file is a download then set the header to passthrough
      if(response.headers["content-disposition"]) {
          res.header("Content-Disposition",response.headers["content-disposition"]);
      }  

[grtjn]

I think I would vote for a more generic solution in which request and response are passed through entirely, maybe even as-is if possible.

I have successfully used the piping features of the 'request' module. Most essential part would be replacing the existing proxy function with:

  function proxy(req, res) {
    var queryString = req.originalUrl.split('?')[1];
    var mlOptions = {
      method: req.method,
      uri: 'http://' + options.mlHost + ':' + options.mlPort + req.path + (queryString ? '?' + queryString : ''),
      headers: req.headers,
      auth: getAuth(options, req.session)
    };
    try {
      //console.log(req.body + '');
      req.pipe(request(mlOptions)).pipe(res);
    } catch (e) {
      console.log(e);
      res.status(500).send('internal error');
    }
  }

[joemfb]

I like the proxy-ing approach as well. Much nicer to avoid the low-level request wrangling the current approach requires ...

[grtjn]

I heard there is a new node module that can do proxying: gulp-webserver. I think it is worth looking into..

[withjam]

Also check out a module 'http-proxy'. I have used it in a couple of projects and swapped it in OBI as well. It's very easy to setup and lets you proxy websockets as well. It looks like this:

var httpProxy = require('http-proxy');
var mlProxy = httpProxy.createProxyServer( { target: dbUrl } );

function proxy(req, res) {
req.headers.Authorization = getAuth(appConfig, req.session);
mlProxy.web(req,res); // this handles the multipart parsing and everything
}

[grtjn]

So as a replacement for 'http' we have:

• request
• gulp-webserver
• http-proxy

Which makes most sense? I believe request was able to replace http entirely. Does that also apply to gulp-webserver and http-proxy? Can anyone tell?

[joemfb]

I used request here: https://github.com/joemfb/mlpm-registry/blob/master/server.js#L67 (see the proxyConfig and proxy functions. I like that flexibility over just using a proxy implementation.

[grtjn]

Somewhat related to #31

[grtjn]

Bitten by the current proxy code again while trying to PUT binary to MarkLogic. I'll move forward to using request. @joemfb I'll take a look at your server.js. Do you think it could be (almost) a drop-in replacement?

[joemfb]

The latest version has been change pretty drastically, but an early version is mostly drop-in: https://github.com/joemfb/mlpm-registry/blob/36430c3207559a32d89cfc2c7669a116c539ebab/server.js

You'll need to remove the route handler for /v1/resources/publish (line 158), and replace application/zip (lines 32, 72) with the mimetype you'll be using for your binary.

[grtjn]

Mimetype? I didn't need that myself. Ah, I guess that was because I got rid of the bodyParser. Didn't see the point of using that.. :)

[grtjn]

Just wanted to mention @withjam created an mlProxy component for this. Maybe ask if that would be easy to integrate, and whether it is available as dependency.

[grtjn]

Looks like this still applies..

[grtjn]

Simplest way out for now is to just copy all headers by default. Looking into that..