awful KIO smb performance / kwallet

[Nicolas Goutte]

On Wednesday 08 March 2006 14:51, David Faure wrote:
> On Wednesday 08 March 2006 13:59, Nicolas Goutte wrote:
> > On Wednesday 08 March 2006 09:55, Martin Koller wrote:
> > > Hi all,
> > >
> > > I tried to copy a lot of files (5263, sum 123,8MB) via kio smb:// from
> > > a server to my local machine (KDE > 3.5.1, Linux).
> > > Effectively the CPU went to 100% and after a while the progress dialog
> > > was still counting the number of files, where I thought what the heck
> > > is going on ...
> >
> > I do not think that this behaviour is SMB-specific and if I remember
> > well, there are already bug reports on that behaiour in KIO.
>
> Listing (recursively) the directories that we're about to copy is no bug,

I meant something like bugs #43356 and #94781
https://bugs.kde.org/show_bug.cgi?id=43356
https://bugs.kde.org/show_bug.cgi?id=94781

> but a necessity.

Sure but if you first gather the files and then only delete/copy/... them, you 
might have lost the directory/file cache.

> But if kio_smb is doing kwallet lookups for every single
> file then that's a bug in kio_smb (or possibly in kpasswdserver? not sure).
> Surely we should reuse the same password for the whole 'session' in the
> ioslave - more precisely, as long as the user+host doesn't change, we don't
> need to look up the password every time.

But the problem is that with KIO slaves, you never know when an operation 
starts and when it ends. So either you keep the "conditions" to be fast and 
you get either compalins than KDE is not secure enough, as it keeps links
or you drop the "conditions" quickly, ut then you have to re-etablish them 
multiple time.

So probably what the KIO slave system miss is a start of startOperation and 
endOperation so that the KIO slave knows when critical data are not needed 
anymore.

Have a nice day!