Re: lwl servers

On Mon, Nov 03, 2003 at 11:43:52PM +0100, Thomas Shaddack wrote:
> 
> > Hmmm, I was just thinking, how hard would be to create a web site based lwl for general users
> > which would depend on something like .cgi for querying for ip address, then setting up a file for
> > users to see as a new page?
> 
> Could be fairly trivial. However, the problem is what to do with
> timeouting of the information; either it would clog the list if it would
> linger forever, or it would have to be periodically refreshed.
> 

If there is someone willing to host a LWL server, it is a much better
thing that what could be kludged up with HTTP.  HTTP could have a
significant problem with proxies (which are very common).  LWL offers
the solution to NAT traversal, and just looking at the code lwld code,
it may already work provided the NAT is friendly enough.  It would be
good to hear of anyone's experiences of lwl under NAT.

The main thing is that the LWL packets can get through the firewall,
which may mean making sure that socket sharing is done there (not
hard, but I haven't checked) The LWLD uses the network address as
reported by the UDP headers, so it should give the right information
for NAT traversal.

An idea that came to me with this talk about LWL servers is to make
use of the SFms RCPT packet type to display a line of advertising
from the LWL line in use - which would then be displayed in
speakfreely.  That way someone supplying an lwl server could see some
return from the investment. 

> 
> Can anyone tell me in brief how exactly LWL works, including the packet
> format, or will it have to wait until I have time to read the source?
> 

You'll probably have to read the source :-(.  The magic seems to be
done in the routine called servicePacket .  It handles RTCP packets,
just adding provided details to a list and querying the list.  There also
seems to be HTML output in there, presumably for use in conjunction
with a web server if desired.

> 
> However, can't we use a different trick? Could a kind of dynamic DNS
> server work? A lot of informations can be put into eg. a TXT record. The
> system could work similar to a cross of a realtime-blackhole DNS service
> with dynamic DNS and reverse DNS lookup; an address of the format
> user.identification.reversednsdomain.com could resolve to its current IP
> and its corresponding TXT record could contain additional information.
> Like eg. TXT records of 4.3.2.1.bl.spamcop.net (where 1.2.3.4 is spammer's
> IP address) do.

I think if someone can host a server, then lwl will do the job, and
it's already there.

> 
> But I think Jabber would be better way to go, as there is a lot of various
> issues (authentication, presence notification, etc.) that are already
> solved there, and connections from the client to the server are always
> outgoing from the client, so there is no problem with NAT in between.
> 

Yes, absolutely.  Except I don't think a standard jabber server will
give all the information needed for NAT traversal of speech packets.
It could be easily customised to do so, but then we still rely on
specific speakfreely jabber servers (unless we can get an approved
jabber extension).  Still, that's no worse than LWL, and it does give
much better authentication.


                      * * *

To unsubscribe from this mailing list, send E-mail containing
the word "unsubscribe" in the message body (*not* as the
Subject) to [email protected]