PATCH: 1st part of the patch to solve the "large file upload problem"

[Dawit A.]

On Friday 06 June 2003 08:03, Thiago Macieira wrote:
> Waldo Bastian wrote:
> >> I presume you are talking about the backwards compatability ctor that
> >> attempts to save the provided buffer from memory to disk.  It was not
> >> intented to be a storage in case the data was re-requested.  The
> >> intention was simply to avoid another duplication (copying) of the data.
> >> However, in hind sight assigning QByteArray's does shallow copying so
> >> there is no reason for it really.  I'll change it.
> >
> >No, I mean you handle this whole resending of data in the job, I think the
> >slave should take care of it.
>
> By doing it this way, the application can send arbitrary data through the
> QIODevice object, not only files. For instance, sockets do work, except
> that you can't rewind them. I don't think we should reject that
> possibility, but that doesn't mean either that the special case of files
> can't be handled specially.

The problem with my approach is that it would work for one action 
KIO::http_post and not another KIO::put because of difference in thier 
design.  I actually tried to handle both cases in the new job, but the 
solution was not clean and really an overkill since the need for re-sending 
data is limited to very few scenarios, such as the need to re-authenticate in 
kio_http.  Anyways, there are other ways to make this work down at the 
io-slave level with little effort.

Regards,
Dawit A.