Re: Download manager and proxies

["Marcelo E. Magallon" <mmagallo debian org>]

On Thu, Dec 02, 2004 at 12:56:03PM +0000, Henry Baldursson wrote:

 > Good point. I'm not so much worried about the authentication, that
 > would be the problem of whichever HTTP library got chosen. Cookies,
 > however, would require more conversation between the broker and the
 > browser. So the browser's acquired cookies could be used to
 > authenticate the browser as a legitimate client for receiving a file
 > (i.e. pron :-). So some callback registering on the part of the
 > caller would be necessary. Bonobo should be equip for that.

 While I agree in general terms, I don't see the need for a callback
 mechanism.  Perhaps I'm missing something related to the finer points
 of cra^Wwebsites like filez.com (or whatever it's actually called).

 When you hand out cookies, you hand out cookies for a given realm,
 which is derived from the URL you are downloading, so the browser knows
 beforehand which cookies it should be handing off to the file broker.

 There are some really sick websites which redirect you a few times
 before the browser can actually start downloading anything.  Here the
 question is when and what is going to be handed off to the file broker.
 You can't offload this task to the user.  Quasi case study:
 SourceForge.  The user sees a "download" hyperlink, if he right-clicks
 + save as he ends up with some webpage in a file (I've done it multiple
 times).  If you just click on the hyperlink, you are taken to a page
 that lists a number of download sites.  If you have a cookie for this,
 you are directed to the same page which eventually redirects you to a
 URL with the actual file.  Which component should be handling all this?
 My choice is the webbrowser up to the point where the actual download
 begins.  That means the browser ends up HEADing a lot, AFAICS.

 Marcelo