Re: Presence service

This is interesting....=0A=0AI think I understand the part about the propag=
ate pipe being well-known so that new peers can announce their presence. Bu=
t how do the other peers reply? Do they use the same progate pipe (I don't =
think so, not sure) or some other well-known pipe?=0A=0AAlso, this takes ca=
re of when a new peer comes online, but what about when a peer goes away? S=
hould peers use the propagate pipe to send a "heartbeat signal" or should p=
eers periodically check in on the peers they know about?=0A=0AThanks,=0AWal=
ter=0A=0A----- Original Message ----=0AFrom: Mark S. Petrovic <mark@petrovi=
c.org>=0ATo: [email protected]=0ASent: Wednesday, December 6, 2006 12:47:20 AM=
=0ASubject: Re: [JXTA user] Presence service=0A=0A=0AOn Tue, Dec 05, 2006 a=
t 05:37:56PM +0100, [email protected] wrote:=0A> Thanks for your response!=
=0A> If I am using the propagate pipe I only need to create it by using the=
 =0A> peer groups pipeservice, there is no need to publish it in the group =
or =0A> create it on the Rdv first, right?=0A=0ACorrect.  What you are desc=
ribing is essentially the technique of using a=0A"well-known PipeID", whose=
 analogue I'm sure you've heard of with things=0Alike SMTP, HTTP, and other=
 important Internet standard services.  There,=0Athey are call well-known p=
ort numbers.=0A=0AWell-known in our case means all peers know the PipeID a =
priori to its=0Ause.  Meaning it is available as an application constant or=
 declaratively=0Aconfigured somewhere in a config file.  There is no point =
publishing it=0Afor the sake of discovery - just hard code it somewhere in =
your app ---=0Abecause you already know you will need it.=0A=0Ahth=0A=0AMar=
k=0A=0A> Vince=0A> =0A> =0A> Mark S. Petrovic wrote:=0A> >Vince, your propa=
gate pipe idea is a good basis for a presence service.=0A> >I travelled the=
 same path you did, first wanting to use advertisements,=0A> >then concludi=
ng that I wanted something more deterministic.  A propagate=0A> >pipe is th=
e price you pay for knowing quickly peers are out there,=0A> >rather than a=
ssuming they are not on some unknown timescale if you cannot=0A> >discover =
their advertisements.=0A> >=0A> >On Tue, Dec 05, 2006 at 11:37:15AM +0100, =
[email protected] wrote:=0A> >  =0A> >>I implemented a presence service for=
 my application a few weeks ago but =0A> >>it is much too slow and unreliab=
le and I want to implement it different =0A> >>this time and would like to =
hear your comments.=0A> >>=0A> >>My current implementation is based on adve=
rtisements that are published =0A> >>every 5 minutes using a timer task suc=
h as Brendon J. Wilson is doing it =0A> >>in his book. Other peers are sear=
ching for these advertisments with a =0A> >>timer task. So the problem with=
 this is that it can take up to a few =0A> >>minutes until a peer is found =
to be online and a peer could go offline =0A> >>and it is not recognized. S=
o I definitely need to change this.=0A> >>=0A> >>My new approach would be a=
 mix of advertisments and a propagate pipe. =0A> >>The advertisment is publ=
ished to the peer group with an application =0A> >>specific description of =
the peer. This advertisment for example holds a =0A> >>pipeadvertisment for=
 a JxtaSocket which is later used to transfer files. =0A> >>When a peer ent=
ers my peer group it searches for the advertisments of =0A> >>other peers s=
o that they can be displayed in the GUI as offline. It also =0A> >>sends a =
"hello" message to all other peers in the peer group with its =0A> >>status=
 such as "online" and gets a response from all other devices that =0A> >>ar=
e online too.=0A> >>=0A> >>Do you have any other ideas on implementing a pr=
esence service?=0A> >>How did other applications solve this?=0A> >>=0A> >>t=
hanks in advance,=0A> >>Vince=0A> >>=0A> >>--------------------------------=
-------------------------------------=0A> >>To unsubscribe, e-mail: user-un=
[email protected]=0A> >>For additional commands, e-mail: [email protected]=
g=0A> >>    =0A> >=0A> >  =0A> =0A> ---------------------------------------=
------------------------------=0A> To unsubscribe, e-mail: user-unsubscribe=
@jxta.org=0A> For additional commands, e-mail: [email protected]=0A=0A-- =
=0AMark S. Petrovic=0APasadena, CA =0Ahttp://www.petrovic.org=0A=0A--------=
-------------------------------------------------------------=0ATo unsubscr=
ibe, e-mail: [email protected]=0AFor additional commands, e-mail: u=
[email protected]=0A=0A=0A =0A_____________________________________________=
_______________________________________=0AAny questions? Get answers on any=
 topic at www.Answers.yahoo.com.  Try it now.