"ravikanth0907@gmail.com" writes:
> I am implementing the IPv6 protocol. I wish to clarify some doubts
> regarding the implementation of it.I have a doubt regarding Source
> address selection algorithm in IPv6. First of all i wish to know, if
> the source address selection has any impact on the routing of the
> packet. What happens if we dont use source address selection algorithm
> in IPv6 and just use one address every time.

It depends.

The most likely effect is that you won't keep up with address changes
in the network. As your deprecated addresses fade away, your
application will fall apart.

> I felt that performing
> source address selection every time when a packet is sent adds
> considerable over head in the critical data path.

I seriously doubt that this matters.

First of all, you don't do source address selection more than once for
connection-oriented protocols, such as TCP. Once set, the source
address can't change for the life of the connection. And if you're
sending bulk data over a protocol not designed for it (such as UDP),
and thus you're worried about the per-packet overhead, then you
probably have bigger problems to contend with anyway.

Secondly, there's no reason that relatively smart hosts can't cache
selection results and internally reuse them when the selection
criteria (the rules and the addresses available) haven't changed.
Thus, as long as you're using a good implementation, it's likely that
it's no more actual overhead than binding a particular address.

But if you're still concerned about it for your application, I would
suggest testing your theory about the expense: try running the
application once without binding the socket, and again with an
explicit bind, and measure the results. Nothing settles such debates
like a good, real-world test.

--
James Carlson, KISS Network
Sun Microsystems / 1 Network Drive 71.232W Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677

wrote:
>
> I am implementing the IPv6 protocol. I wish to clarify some doubts
> regarding the implementation of it.I have a doubt regarding Source
> address selection algorithm in IPv6. First of all i wish to know, if
> the source address selection has any impact on the routing of the
> packet. What happens if we dont use source address selection algorithm
> in IPv6 and just use one address every time. I felt that performing
> source address selection every time when a packet is sent adds
> considerable over head in the critical data path.

I think the problem is that IPv6 assigns multiple IP addresses to each
interface, but they do not all have the same scope. An extreme example,
you wouldn't use a link-local source address for packets destined to the
Internet. But there are less extreme examples, where a given source
address may not be reachable from the destination network at which
you're trying to open a session.

RFC 3484 discusses these issues.

Bert

Thank you james..

James Carlson wrote:
> "ravikanth0907@gmail.com" writes:
> > I am implementing the IPv6 protocol. I wish to clarify some doubts
> > regarding the implementation of it.I have a doubt regarding Source
> > address selection algorithm in IPv6. First of all i wish to know, if
> > the source address selection has any impact on the routing of the
> > packet. What happens if we dont use source address selection algorithm
> > in IPv6 and just use one address every time.
>
> It depends.
>
> The most likely effect is that you won't keep up with address changes
> in the network. As your deprecated addresses fade away, your
> application will fall apart.
>
> > I felt that performing
> > source address selection every time when a packet is sent adds
> > considerable over head in the critical data path.
>
> I seriously doubt that this matters.
>
> First of all, you don't do source address selection more than once for
> connection-oriented protocols, such as TCP. Once set, the source
> address can't change for the life of the connection. And if you're
> sending bulk data over a protocol not designed for it (such as UDP),
> and thus you're worried about the per-packet overhead, then you
> probably have bigger problems to contend with anyway.
>
> Secondly, there's no reason that relatively smart hosts can't cache
> selection results and internally reuse them when the selection
> criteria (the rules and the addresses available) haven't changed.
> Thus, as long as you're using a good implementation, it's likely that
> it's no more actual overhead than binding a particular address.
>
> But if you're still concerned about it for your application, I would
> suggest testing your theory about the expense: try running the
> application once without binding the socket, and again with an
> explicit bind, and measure the results. Nothing settles such debates
> like a good, real-world test.
>
> --
> James Carlson, KISS Network
> Sun Microsystems / 1 Network Drive 71.232W Vox +1 781 442 2084
> MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677

In article <1155024998.015631.88440@m73g2000cwd.googlegroups.c om>,
"ravikanth0907@gmail.com" wrote:

> Thank you james..

Was it really necessary to quote 100 lines just for this?

--
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***
*** PLEASE don't copy me on replies, I'll read them in the group ***