Re: [Paperclips-discuss] HEAD, was: API_23_ALPHA_2_1

[Nic Ferrier]

>>> "Christopher K. St. John" <address@hidden> 25-Jun-01
11:09:43 PM >>>

   Speaking of which:
   1) www.java: doGet():
     target.connect();
     if(!target.getContentType().equals("text/directorylisting"))
  
    public String URLConnection.getContentType()
      Returns: 
            the content type of the resource that the URL
            references, or null if not known. 

    the 'null if not know' part is the problem, it means
    that paperclips won't serve .tgz for instance. (I have a
    really big tarball I was using to test download speed on
    NbioServer, so I decided to try it out on paperclips...)
    Easy enough to fix, but is that code scheduled for a
    rewrite? Is it worth cleaning it up?

Yes, it's worth cleaning up. I'm not going to change that code, it
works pretty well.

The problem is really with the getContentType method because it
should handle mime type mapping better. I'm not really sure how to do
this though... the API specs the registration of mime types in a DD
but might it be a good idea for Paperclips to read a mime.types file
in some default places if the DD doesn't register any mime types?

I have some mime.type parsing code in the Classpathx project so it
should be easy to do.

Let me know what you think.


    2) Using httperf[1] hangs things pretty quickly: the response
    times gradually increase until everything appears to stop.
    httperf (and apache's ab) are both pretty ugly with what
    they send over the network, but hanging is probably not
    ok. (It came up because I was using httperf for testing
    NbioServer, so I decided to try it out on paperclips ...)

No, that definitely doesn't sound right...

I have load tested previous versions of Paperclips and it worked okay
up to 1000 connections (and that's all I needed so I stopped).

Can you post the test procedure for httperf? I don't anything about
it.


   Should I be using the savannah bug tracking stuff?
   I didn't see a link from the paperclips page and I 
   assumed it wasn't turned on.

I'm not sure... I've always fixed bugs via email before. That is the
current GNU standard.

Is anybody else keen to use the support-manager stuff?


Nic