Joerg Sommer <[EMAIL PROTECTED]> writes:
> begin Andrew M. Bishop <[EMAIL PROTECTED]> wrote:
> > prefers chunked encoding since it serves so much dynamic content. The
> > proxy might not care (WWWOFFLE) or it might prefer Content-Length
> > (Polipo). Who knows what the browser prefers, if it is parsing the
> > content into markup tags then it probably doesn't care too much as
> > long as there is a stream of bytes.
>
> You are right in the case of markup data. But it's a difference, if you
> download a binary object like a pdf file.
I was talking about what the browser program prefers, not what the
user of the browser prefers. I don't think that there are any cases
where the brower needs to allocate large data structures for the raw
data that arrives (except perhaps the in-memory cache, but I don't
know that raw data is stored there). If the data is HTML it is parsed
for tags, if it is an image it is passed to the image library as a
data stream, if it is for an external viewer it is saved to a
temporary file.
> When you download a pdf or a
> tar.gz and you see after two minutes it is still downloaded, but you
> don't know how long, it's frustrating. You don't know is the file 2.4MB
> and you should wait the 10 seconds or is the file 12MB and you can abort
> the download.
I can see that it might be frustrating in some cases. Since I don't
often use WWWOFFLE in online mode I don't see this often.
Since it is a feature that is causing concern for many people I will
look at adding back in some Content-Length information for the cases
where I can.
> And the point is, wwwoffle shouldn't modify binary data. So
> I don't see any reason, why wwwoffle should remove the C-L header on
> binary data. Please, let don't remove the header on binary data.
WWWOFFLE doesn't know which files are binary and which are not. The
reason that WWWOFFLE removes the header is that there are many special
cases when it must remove the header or there is not a header. To
*never* send out a Content-Length header means that WWWOFFLE can never
send out an incorrect one, or send one when it shouldn't have done.
An incorrect header is worse than no header since it will confuse the
client but all clients will accept there not being a header.
--
Andrew.
----------------------------------------------------------------------
Andrew M. Bishop [EMAIL PROTECTED]
http://www.gedanken.demon.co.uk/
WWWOFFLE users page:
http://www.gedanken.demon.co.uk/wwwoffle/version-2.8/user.html
