Sebastian Werner <[EMAIL PROTECTED]> writes:

> Am 18.05.2007 um 14:11 schrieb [EMAIL PROTECTED]:
>
>> Sebastian Werner <[EMAIL PROTECTED]> writes:
>>
>>> I don't like the solution to be integrated in qx.core.Target. I would
>>> like it better to add specific events to a special event router which
>>> then informs registered objects about any of these events. This is a
>>> lot cleaner than exclude tables and a deep integration in
>>> core.Target. The overhead of many events would be to dramatic. I can
>>> imagine of something like a widget monitor which reacts on all
>>> widgets relevant stuff and for example a IO monitor which monitors
>>> all IO related things.
>>>
>>> What do you think?
>>
>> Hi Sebastian,
>>
>> That concept sounds fine.  I'm not sure I follow where you'd add  it, though,
>> if not in Target.  If one wants to monitor all, or some very large  subset of
>> events, where else could you put it that has access to the  information?
>>
>> I expect that in the implementation I checked in last night, the  overhead is
>> quite small since it effects only events with listeners if the  variant isn't
>> enabled.  I am concerned about the overhead when the variant _is_  enabled.
>
> Derrell,
>
> Each overhead even if really really little is to much in my opinion  because
> it also sits on time critical stuff.

That may be.  Part of my testing was going to be to determine how much
overhead there was in the simple "if" condition (with map look-up for
exclusions).

> The whole idea for the message bus is instead to use it only for not-so time
> crititcal information.
>
> I would say, that a first step, before implementing something is, to  get a
> clear view of which currently existing events are interesting.  I think a
> whitelist is better than a blacklist in this case. A  message bus listener
> which reacts on all the possible events of  different areas e.g. io, widgets,
> focus, whatever is not quite often  - if it exist at all.

The problem with a whitelist is that one should be able to use a monitor to
determine what events are being dispatched.  In other words, we don't always
know which events are interesting ahead of time.  To set up the whitelist, you
have to already know, so that pretty much defeats that particular purpose.  It
should be possible to get notified of *all* events (even if that incurs some
overhead).  If the overhead is significant, the monitoring should be enabled
by a variant, so apps that don't require the monitoring can run at full speed.

>> I'd definitely like to hear more about how you would implement the  special
>> event router.  Sounds quite interesting!
>
> OK, lets continue our discussion on Monday.

Great.  Christian, I'd suggest not implementing anything based on what I
checked in last night.  I suspect it'll get reverted, and replaced by
something completely different.

Cheers,

Derrell

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
qooxdoo-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/qooxdoo-devel

Reply via email to