RE: Developing and Testing panel applets

[Michael Meeks <michael ximian com>]

Hi Sergey,

On Sat, 2002-12-21 at 19:48, Sergey V. Oudaltsov wrote:
> > 	+ 1% of users use JimBobHackedLeetShLibApplet which corrupts
> > 	  memory badly, uses global variables, can't cope with multiple
> > 	  instances etc.
>
> :) BTW, what would be the way to PREVENT the second instance of the applet?
> I would really need this. Not related to the panel itself - it is just very
> difficult (impossible?) to design xkb switcher in a way which would support
> multiple instances.

	Run the applet in a separate process; then it is at least unique
per-screen [ and with a little more work in 2.4 (or whatever) we should
be able to tweak that in the .server files to unique / per-screen /
per-display / per-foo etc. ].

	A separate process applet is really the best decision for lots of cases
- particularly where you want a unique server.

> OK. That is the kind of a statement I expected. Probably I should make my
> applet back to app, how do you think?

	I agree; it's not that hard either which is good.

> BTW, any idea: what it a price of a process in Linux (RAM-wise, in kernel)?
> And whether in-process CORBA server works any faster then out-process...?

	In-process CORBA calls are almost 0 overhead; comperable to a
dynamically bound call or three. Out of process it depends per platform
on things like the cost of context switching, how many calls you do, how
well designed the interfaces are etc. For an applet you should notice no
performance problems however.

	HTH,

		Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot