[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: KIOSlaves and configuration
From:       Sergio Moretti <sermore () libero ! it>
Date:       2001-02-06 21:33:25
[Download RAW message or body]

On Tuesday 06 February 2001 16:49, David Faure wrote:
> Let's merge and restart this discussion from scratch :)
>
> My preferred approach would that the slaves would _ask_ for specific
> configuration values, when they need them. They currently ask that
> directly from the config files, using KProtocolManager, which makes
> it very difficult to make application-dependant (the slaves have no
> idea about which app they relate to). So my idea would be that they
> ask for configuration values via the SlaveBase->SlaveInterface pipe,
> i.e. they would ask the application.

so you are thinking about making some blocking operations like the 
openPassDlg, one for each conf entry? ok, but I think that could exists 
2 classes of configurations, one that is needed for every operation 
that a slave has to do (for example Agent header for an http request), 
for which is better to pass it at the beginning of the slave-job 
interaction, and a class of rarely used confs, maybe like the password, 
for which your solution fits well. Anyway the "interactive" solution is 
needed for unexpected events, so, if I have to choose between the two, 
my choice is for this. The drawback is that you add synchronization 
points between slave and job, and maybe you stop the slave to wait for 
5 times instead of sending 50 bytes at the beginning.

> In the application, a KProtocolManager instance would be created with
> initially the desktop defaults, but the class would accept "runtime
> modifications", i.e. the application could set values in that class
> _without_ that value being saved as the global default. I guess we
> would keep the static methods for writing the global defaults, but
> have an instance of the class for the application-specific values.

Yes, maybe making a read-only interface divided from the read-write 
one. BTW if you take off this from slave, don't you lose the automatic 
update of the defaults? 

> If we go for this solution, and if nobody else implements it, I will,
> but I'd really prefer to concentrate on other things after 2.1 if
> possible, so patches will be welcome :)

I'd like to contribute, but I need help and time, so if you have both...

cheers,
	Sergio

[prev in list] [next in list] [prev in thread] [next in thread]