Ann, David & Sanjiva, I kept quiet all this time bcos, I was the only person who voiced my concern about this issue a while back. I will breifly outline my concerns for the benifit of thers (Ann has already responded to my comments. Whether I agree with them or not is a different thing)
a) IMHO if the context heirachy restoration is not brought back to the original state then whats the point of serializing the context heirachy in the first place. b) As Ann mentioned before the idea or the value of the proposal "is to restart a message in progress from where it was stopped w/o having to start from the begning". IMHO this is not much of a value addition to axis2 compared to the complications in the code base required to achive this. Here is my point. We are spending far more processing power to save the message contexts anticipating a server crash, than to have the aborted messages start from the beging if a server crash occurred. If u pair it up with Sandesha2 we are garunteeing the excution of each message, so we are not loosing any messages, but will restart processing from the begining if a server crash occured. c) Sorry for being blunt, but I honestly think, that this proposal is being pushed so hard bcos it works with some IBM internal product. I actually don't care if this improves axis2 in some way or the other. I think the complications introduced to the core far out weighs the benifits provided by this implementation. Regards, Rajith Btw, David, Can u please outline your concerns regarding clustering ? I may have missed this. Also I would like if u can explain the following comment a bit more.
I certainly don't believe that the clustering proposal had all the answers or gave a view of an implementation that I want in Axis2.
On 12/18/06, Matthew Lovett <[EMAIL PROTECTED]> wrote:
Hi Sanjiva, Sanjiva Weerawarana <[EMAIL PROTECTED]> wrote on 18/12/2006 14:49:15: > > OK how about this- Ann, since you just want to use this feature for your > own persistence model for Sandesha and nothing else, then you don't need > to put this code in the core at all: just write the same writeObject and > readObject methods in a utility class in Sandesha2 and call it yourself. > The only diff is where the code goes. I don't think that we can implement this in the Sandesha layer. There is too much private state inside the Axis objects that needs to be handled, and simply making all the objects have no-arg constructors & get/set for everything would be very ugly - and break encapsulation. That's why the classes themselves need to implement their own save/restore logic. > I'm surprised I'm the only one who cares about this to comment! > I do care about this stuff too, and I'd like to see it in Axis. Matt --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
