Hello,

I'm writing a C++ software which send data to a second computer through
a UDP connection.
What I would like to do is to oblige my connection to pass by an
intermediate server between these two computers. Is it possible? If
yes, how please?

Best Regards,
Rod

In article <1160866762.813902.224710@f16g2000cwb.googlegroups .com>,
<rodrigue.roland@gmail.com> wrote:

>I'm writing a C++ software which send data to a second computer through
>a UDP connection.
>What I would like to do is to oblige my connection to pass by an
>intermediate server between these two computers. Is it possible? If
>yes, how please?

No, there isn't any way to do that from your end.

The closest you can get is to use "Loose Source Routing"
or "Strict Source Routing"

http://www.synacklabs.net/OOB/LSR.html

However, it is enough of a security problem that most networks
will disable recognition of that option, or drop the packets entirely.

Hello,

rodrigue.roland@gmail.com a écrit :
>
> I'm writing a C++ software which send data to a second computer through
> a UDP connection.
> What I would like to do is to oblige my connection to pass by an
> intermediate server between these two computers.

An intermediate *server* or *router* ?

Hello,

Thank you for your response...
I've tried to understand this technique in TCP/IP illustrated vol.2.
In fact, that was just an idea to ameliorate "hole punching" method.

If two clients, A and B, try to connect themselves through my server (a
sort of proxy), I could known exactly which ports are used at each NAPT
(it's the main problem of hole punching).

When they'll be connected together, I could change the route of
connection and remove my server (cf. ch 20 of TCP/IP illustrated vol
2)...

It's just an idea!

Best Regards,
Rod

uPnP could be also very important for "hole punching" for home
networks. It permits to programmatically control a UPnP-Capable
Firewall/NAT. What do you think about it, is it useful?

A VB source code:
http://www.knoxscape.com/Upnp/NAT.htm

In article <1161020832.888114.230780@k70g2000cwa.googlegroups .com>,
<rodrigue.roland@gmail.com> wrote:

>If two clients, A and B, try to connect themselves through my server (a
>sort of proxy), I could known exactly which ports are used at each NAPT
>(it's the main problem of hole punching).

>When they'll be connected together, I could change the route of
>connection and remove my server (cf. ch 20 of TCP/IP illustrated vol
>2)...

I think we might have a subtle English usage that is interfering with
being sure we understand the question. I will explain by analogy.

Between my work and my residence there is a river. There are two
convenient bridges over that river. One of the two bridges mostly
has wide roads and houses on both sides of the bridge, but one
side of the other bridge has narrower roads and a series of shops
and restaurants, and traffic there is slower because the roads
are narrower and there are many people going to the shops or
restaurants.

When I go from work to home, I can travel by way of either bridge.
If I want to, I can "pass by" the busy area. In English usage,
that means that I *do* go by the busy area. If I go the other, less
busy way, then in English usage, I "bypass" the busy area.

In English, to "pass by" something always means to travel near it,
but not necessarily to stop there. I could go home and "pass by"
way of one of the restaurants and stop and pick up something to take
home to eat, or I could go home on the same road and "pass by"
the same restaurant meaning that I went near it but did not stop.
So "pass by", without extra words, is not clear as to whether you
stopped or not -- you at least went close to it.

In English, to "bypass" something means to avoid it or go around it:
I can "bypass" the heavier traffic of the commercial district by taking
the other bridge. Also, to "bypass" something can mean to
prevent it from working properly, or to remove it from a circuit, or to
add an extra circuit so that the item is not triggered or not involved
with something; for example, you can "bypass" an alarm system. If
something is bypassed, an alarm or circuit or road, then it is not
involved with what you are doing, or is prevented from being involved
with what you are doing.


Now, your original posting asked to have packets travel and "pass by"
a particular server. In English that would normally be read as
indicating that you want to force the packets to travel by way of
that server, like you "passed by" the store meaning you
went by way of the store buy some bread there. My response about
"source routing" was telling you that forcing a packet to travel
through a particular server is usually not allowed on networks.

In your follow-up response to the other poster, it sounds as if you
have two systems, A and B, that are both talking to a common server, S,
and you want to have A and B talk to each other directly, -not- stopping
at S. In English, that would be having the traffic between A and B
"bypassing" the server S -- that S would not get involved, and
possibly that the traffic would not go anywhere close to S.

Could you clarify whether you want the traffic to "pass by" the server,
meaning you want the traffic to go through the server, or if you
want the traffic to "bypass" the server, meaning that you do NOT want
the traffic to go through the server?

rodrigue.roland@gmail.com a écrit :
>
> I've tried to understand this technique in TCP/IP illustrated vol.2.
> In fact, that was just an idea to ameliorate "hole punching" method.

Sorry, I don't know about this book neither "hole punching".

> If two clients, A and B, try to connect themselves through my server (a
> sort of proxy), I could known exactly which ports are used at each NAPT
> (it's the main problem of hole punching).
>
> When they'll be connected together, I could change the route of
> connection and remove my server (cf. ch 20 of TCP/IP illustrated vol
> 2)...

Are you talking about someting like RFC 3489 - STUN - Simple Traversal
of User Datagram Protocol (UDP) Through Network Address Translators (NATs) ?