You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@avalon.apache.org by Anton Tagunov <at...@mail.cnt.ru> on 2003/05/13 15:22:37 UTC
Re[2]: Future Direction for Avalon
Hello, Leo and All!
>> From: Niclas Hedhman [mailto:niclas@hedhman.org]
>>
>> Sharing of components between containers will introduce very
>> interesting classloading issues, if you intend to support
>> reloadable containers and/or components.
LS> Actually, I think the Fortress/ECM usage pattern ...
LS> .. for each component that is to be reloaded, we can block
LS> lookup()s and only allow release(). When all instances
LS> are released - reload.
Have been thinking much (webapp background) what if
- old clients continue using the old class
- new clients lookup and get instances of the new class?
As soon as the impl is reloaded and not the api it should be ok.
<disclaimer>
I'm an outsider without proper understanding
</disclaimer>
- Anton
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@avalon.apache.org
For additional commands, e-mail: dev-help@avalon.apache.org
Re: Future Direction for Avalon
Posted by Niclas Hedhman <ni...@hedhman.org>.
On Tuesday 13 May 2003 05:22 am, Anton Tagunov wrote:
> Hello, Leo and All!
>
> >> From: Niclas Hedhman [mailto:niclas@hedhman.org]
> >>
> >> Sharing of components between containers will introduce very
> >> interesting classloading issues, if you intend to support
> >> reloadable containers and/or components.
>
> LS> Actually, I think the Fortress/ECM usage pattern ...
> LS> .. for each component that is to be reloaded, we can block
> LS> lookup()s and only allow release(). When all instances
> LS> are released - reload.
>
> Have been thinking much (webapp background) what if
> - old clients continue using the old class
> - new clients lookup and get instances of the new class?
This may not be good enough, since the "old client" may be a never-stopping
application, but that needs the upgraded service.
I think 2 things is required;
1. A defined "availability (notification) protocol", which is the underlying
driving mechanism for making services available or not.
2. A higher level system (someone called it interceptors) that understands the
availability protocol, and makes it easy for the "client" to access the
service, without worrying about the "availability (notification) protocol".
Niclas
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@avalon.apache.org
For additional commands, e-mail: dev-help@avalon.apache.org