Re: [Bug-wget] Truncated files ... --ignore-length bug ??

[Daniel Stenberg]

On Wed, 12 Aug 2015, Darshit Shah wrote:

> The server responds with a HTTP 200 response, so it is not returning only 
> partial content / attempting to stream. The way Wget reads that response is 
> that the server has sent all the data it had.

This server is clearly not behaving correctly. It states clearly what the size 
of the file is and both wget and curl then stops downloading after that size. 
This is clearly following the HTTP 1.1 spec.

> Also, since curl too gives you the same problem, but Firefox does not, let
> me ask you one question, Did you load the same session cookies on Firefox
> too? Or did you authenticate and start a new session?

My guess is that the answer is much more complicated than so. First we didn't 
get which Firefox version that was used, which is highly relevant since I've 
been on a mission during the last year or so to straighten up Firefox's HTTP 
framing checks 
(http://daniel.haxx.se/blog/2015/03/02/more-http-framing-attempts/) so that it 
becomes more strict and less forgiving about mismatches in the content-length 
header and what is actually sent.

Still, even today with the very latest version, Firefox is more forgiving (or 
spineless) than wget and curl when it comes to these details. Primarily 
because the Internet is populated with too many of these errors and the other 
browsers already swallow these vioations fine so Firefox cannot throw errors 
to the users on them as that would only make users switch to one of the 
browsers that does not.

My blog and Firefox bugzilla have lots of more details on my futile fight in 
this area.

--

 / daniel.haxx.se