[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [xougen] Re: xouvert-general Digest, Vol 1, Issue 13

From: weigelt
Subject: Re: [xougen] Re: xouvert-general Digest, Vol 1, Issue 13
Date: Wed, 27 Aug 2003 21:53:38 +0200
User-agent: Mutt/1.3.27i

On Wed, Aug 27, 2003 at 12:38:29PM +0200, >> G-LiTe / wrote:

> I wouldn't know exactly about that. You could write an extension that
> dynamically links to the libraries and just receives function calls over
> the network to execute them. Sounds sortof like RPC. If we did that we'd
> need some serious restrictions on the extension for security reasons.
ACLs based on procedure namespaces/groups ?

i.e. you can write down:

DRI:    *                               ==> deny
        source.type=local               ==> allow

Widget: *                               ==> allow
        source.host IN broken_clients   ==> deny

> On top of that, clients may not know exactly what the window looks like at
> some point and you still have to send an event every time the mouse hovers
> over or clicks a button.
Eh ?! Why ?
When an client uses the Widget extension, it creates some widgets and
then tells the server what to paint in several situations (in case of an
button: normal,focused,mouse-over, etc). This can be easily done with
simple property lists.

> Let's see what else I can think of ;) ... the library itself would have to
> support the extension and check the version of it on the server too. All
> necessary functions in the library should be wrapped to the server. There
> should be some extra functionality in the extension to get a bitmap of an
> area so clients know how to, for example, invert certain parts of the
> window, which is required for selections and the likes.
It is possible to grab some bitmap. But why not an InvertRegion function 
for this stuff ? Do as much as we can (and is useful) in the display server.

> Libraries like gtk (actually glib) have an event loop alot of applications
> use to implement their own events too. The extension should be able to send
> function calls back. How are we going to get the pointers across anyways?
No pointers. Messages can be listened. Each listener has an name or an
ID or something like that. This one is handled by the client stuff.

> It all sounds a bit hackish to me. It was an okay idea but it has it's
> drawbacks. In my opinion, having the toolkit run server side really doesn't
> have that many advantages. Does the bandwidth usage really matter that much
> over a local socket or LAN? 
It _really_ matters in low-bandwith environments. Not evryone has an
10mbit link coming out of the wall as water or electricity.

> I can't really imagine anything working smooth over a WAN and if people 
> want to use it, I'd suggest just using vnc and allow the application to 
> read required information over a socket from the computer it was actually 
> suppose to run on.
Even with RDF+friends it is really too slow. Well, its the wrong concept, 
to grab an Image and then look how we can compress. The right way is to 
work on the highest level which is possible.

Look at Citrix Metaframe. They somewhere intercept the windows GUI stuff 
and only post the really necessary data from the GUI API calls over the net. 
And so it really becomes _very_ fast.

 Enrico Weigelt    ==   metux ITS 
 Webhosting ab 5 EUR/Monat.          UUCP, rawIP und vieles mehr.

 phone:     +49 36207 519931         www:       http://www.metux.de/     
 fax:       +49 36207 519932         email:     address@hidden
 cellphone: +49 174 7066481          
 Diese Mail wurde mit UUCP versandt.      http://www.metux.de/uucp/

reply via email to

[Prev in Thread] Current Thread [Next in Thread]