[Seaside] VW questions
keith_hodges at yahoo.co.uk
Tue Oct 30 01:04:41 UTC 2007
> It's a bad idea because it makes components that aren't reusable,
> theybecome very tightly bound to their parents.
I saw that on a blog somewhere, but I am not sure I buy it. Surely they
only become bound to their parents if you use the relationship over rigidly.
Consider the situation anthropomorphically... If I am presently in a
car, I can see and know that I am in a car. If I choose to shout then
all of the car can hear me. If I am in a bus, I can see I am in a bus,
and again if I shout in the bus the bus driver can hear me. However I am
not chained to the bus and can take the train if I want to.
When modelling physical telecomunications equipment I always implemented
parent-child relationships, and certainly never regretted it. Basically
I used the #parent for routing semantic messages.
When a card was removed from the rack it would instanciate a
SemanticMessage-cardRemoved: aCard (based upon the ideas of
SemanticMessages in SmalltalkAgents) This message would move up the
heirarchy looking for an object to either a) directly handle the message
#cardRemoved: aCard or if the item had any subscriptions registered such
that on: selector send: message to: subscriber.
When the Alarm card was inserted in the rack it would register for
cardRemoved: events and raise the appropriate events.
theRack on: #cardRemoved: send: #cardRemovedAlarm: to: self.
Subscriptions could also add arguments to a message as it passed through.
I am begining to think that I might reimplement self routing
SemanticMessages with a slightly more announcement like flavour for Seaside.
>> Announcements seems limited to me since if using a global
>> announcer. I cannot see how two similar components on one
>> page, each trying to announce something to their
>> parents/containers would be able to.
> You don't have to use a global announcer, every component can have it's own
> announcer and you can have parents subscribe to events their children throw
> that they're interested in, this keeps the children loosely coupled and
And this was what led me to wondering how on earth this works at all.
How do the children find the parents announcer, if they do not know whom
their parent is?
I think that I am struggling to understand announcements and I cant find
the original articles any longer.
> You could also just have a sender on the announcement to let the parent
> decide if the sender was a child, if you prefer a single global announcer.
>> any thoughts would be appreciated.
> Every time I've gone down the parent path, I've regretted it. I'll stick
> with announcements, much better approach.
> Ramon Leon
Thanks very much for your prompt reply
More information about the seaside