Many thanks BlueRaja,
So far I had implemented your second, very valid I would say,
approach. Just a question: the communication between nodes in P2P is
always based on TCP connections? Could UDP ones get involved? This
question raised actually the problem of how a node realises a
neighbour has failed. If it uses TCP it is more or less obvious how.
But if not? Is there such a system?
Thanks again for your help,
-- George
On Jun 20, 10:32 pm, BlueRaja <blueraja.ad...@gmail.com> wrote:
> In an actual P2P application, each client would have a unique IP address.
> Although you may indeed use these as IDs (perhaps, for example, as keys to a
> large hash-table), it would probably be simpler to just use the pointers as
> the "ID's" to connect one node to another, internally. When displaying the
> connections to the user, however, you would probably want to use the IPs.
> As for deleting inactive nodes when they are no longer needed, the simplest
> way would be to check all the nodes that point to the current node (since it
> is a P2P application, the graph should be bidirectional), remove all
> references to the current node, and then deallocate it (this would be
> partially analogous to a real-world socket closing when it finds that the ip
> is was connected to is no longer active). However, if you wanted an even
> more "real-world"-like approach, you could give each node a boolean 'active'
> or 'live' property: when a live node wants to contact another node, it will
> check the other node's live status it; when the live node finds that the
> node it's looking for is dead (ie. inactive, offline or whathave you), it
> will remove it from its "contact-list" (ie. the edge-list. This is again
> analogous to a socket closing). Using this approach, each node could have a
> reference counter that keeps track of how many other nodes are pointing to
> it - when that counter reaches 0, the node can safely be deleted (this is
> the approach that many naive garbage collectors use).
>
> Hope that didn't sound *too* rushed :)
> -BlueRaja
>
> On 6/20/07, gexarchakos <gexarcha...@gmail.com> wrote:
>
>
>
> > Hi BlueRaja,
>
> > Well, the thing is that I am building a P2P simulator. In P2Ps there
> > is not guarrantee that a node has always valid and active neighbours.
> > A node may disappear without notifying its neighbours. Therefore,
> > explicitly deleting all the pointers to a deleted object does not 100%
> > represent an actual P2P.
>
> > -- George
>
> > On Jun 20, 3:08 pm, BlueRaja <blueraja.ad...@gmail.com> wrote:
> > > If there is any possibility that there will still be references to an
> > > object elsewhere after it is deleted, it should not be deleted in the
> > first
> > > place; this is not a problem with pointers but rather with the design of
> > > your program.
> > > You would have to give a more in-depth description of the program
> > (and,
> > > especially, what should happen when a node is deleted) in order for us
> > to
> > > give a recommendation of what to do with your garbage-pointers.
>
> > > -BlueRaja
>
> > > On 6/20/07, Saurabh Saha <saurs...@gmail.com> wrote:
>
> > > > Hi George
>
> > > > I completely agree with you but the doubt that I possess is how do you
> > > > plan to allocate the same memory area that you freed because as soon
> > as you
> > > > free taht portion of the area, the area becomes a part of the free
> > pool of
> > > > memory space that is available to be allocated with of course the
> > garbage
> > > > values or values wrt the earlier object.
>
> > > > Now what exactly is the probability that a new node will take up
> > exactly
> > > > the same area that you freed in the last node as now it is not about
> > the
> > > > memory area but the picture is rather the free pool of memory in the
> > heap.
>
> > > > Now allocation of memory to the new node is the OS's responsibility
> > and
> > > > hence it takes care of it not the coder at the code level.Wat say????
>
> > > > Regards
> > > > Saurabh
>
> > > > On 6/20/07, gexarchakos <gexarcha...@gmail.com> wrote:
>
> > > > > Many thanks Saurabh,
>
> > > > > actually you are right, that's exactly what I meant. I think I agree
> > > > > with your points...
>
> > > > > I am building a network simulator and created a directional graph
> > and
> > > > > initially I was using the pointers as IDs. But now I expand the
> > > > > simulator with dynamic insertion and deletion of nodes. This means
> > > > > that a new node may take the same memory space as an old deleted
> > one.
> > > > > In this case, not updated pointers from other nodes to the old one
> > > > > will actually point to the new node. If no new node is inserted in
> > the
> > > > > same memory space then these pointers will be broken, right?
>
> > > > > That was the problem that generated the question.
>
> > > > > Many thanks again,
> > > > > --George
>
> > > > > On Jun 20, 11:54 am, "Saurabh Saha" < saurs...@gmail.com> wrote:
> > > > > > Hi George
>
> > > > > > When u talk about deleting a memory address what exactly do u
> > actually
> > > > > > mean-It does not mean u delete any memory area but rather it means
> > > > > that u
> > > > > > free portion of d dynamic memory which was allocated to the object
> > at
> > > > > > runtime so I believe there is of course a possibility that the
> > same
> > > > > memory
> > > > > > area can be taken up by another object unless of course the
> > developer
> > > > > > forgets to de allocate the memory allocated thus resulting in a
> > memory
> > > > > leak.
>
> > > > > > So talking about delete from the OS point of view does not quite
> > mean
> > > > > that
> > > > > > whatever u delete from th UI is actually deleted but rather that
> > the
> > > > > > reference entity pointing to d memory area gets allocated 2 Null
> > > > > permanently
> > > > > > and hence cannot point to d memory area it was earlier pointing
> > to.
>
> > > > > > Correct me if i am wrong guys,.
>
> > > > > > Regards
> > > > > > Saurabh
>
> > > > > > On 6/20/07, gexarchakos <gexarcha...@gmail.com> wrote:
>
> > > > > > > Hello everyone,
>
> > > > > > > just a question on pointers:
> > > > > > > Assume that you dynamically allocate memory for an object.
> > Sometime
> > > > > > > later, you dynamically delete it...
> > > > > > > Is there a possibility that another dynamically created object
> > will
> > > > > > > take the same memory address as one previously deleted? In that
> > > > > case,
> > > > > > > pointers cannot work as IDs of objects, right?
>
> > > > > > > To me it makes sense if there is such a possibility. I'd like
> > your
> > > > > > > opinion...
>
> > > > > > > Thanks,
> > > > > > > -- George
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "C++ Programming" group.
To post to this group, send email to Cpp-Programming@googlegroups.com
To unsubscribe from this group, send email to Cpp-Programming-unsubscribe@googlegroups.com
For more options, visit this group at http://groups.google.com/group/Cpp-Programming?hl=en
-~----------~----~----~----~------~----~------~--~---
