RFC: KIO::Scheduler changes... [PATCH included]

[Dawit A.]

On Monday 14 September 2009 06:57:11 David Faure wrote:

> I object to the job.cpp changes for the reasons indicated in the earlier
> thread about this (it would potentially deadlock file-management operations
> like KIO::file_copy), so I'm glad you're taking these out.

Right, because I finally understood what you were saying about the dead locks 
in the earlier thread...

So now I too object to my own patch! There is nothing that prevents the 
scenario explained in the earlier thread from happening when one attempts to 
download something from an http server as well!

> We still need to find a solution for the "copying lots of files" and
> "connecting to FTP servers multiple times", but it seems this needs to be
> done at a higher level than your changes, so your changes are really only
> for webbrowsing. 

Yep... I guess some sort of request queueing is needed somewhere other than 
KIO::Scheduler...

> But then, was there really a problem of "too many
> connections to the same server" in the case of webbrowsing?

No, there was not. The problem for me was way too many ioslaves when visiting 
any moderately complex site. For example, when one visits dot.kde.org, 
depending on your connection speed, something like 28 instances of kio_http 
are created to fullfill the job requests...

> I thought khtml's own scheduler (loader) took care of that; dunno about
> kdewebkit though.

I guess that is what khtml does because it does not suffer from this problem 
and it seems that is what I have to do for kdewebkit as well.