[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kfm-devel
Subject:    Re: Fwd: SRV support in konqueror ?
From:       Waldo Bastian <bastian () kde ! org>
Date:       2001-07-24 16:52:29
[Download RAW message or body]

We pretty much depend on the operating system to do the DNS for us. We use 
getaddrinfo(3) to do name lookups. getaddrinfo is part of glibc and I assume 
that support for SRV should be added there.

Thiago is our DNS expert though, maybe he can give you a better answer.

See http://sources.redhat.com/glibc/ for how to reach the glibc developers.

Cheers,
Waldo

On Tuesday 24 July 2001 08:45 am, David Faure wrote:
> Can someone answer this ? I have no clue about this stuff.
>
> ----- Forwarded message from Peter Håkanson <peter@ipsec.nu> -----
>
> Date: Tue, 24 Jul 2001 14:40:10 +0200 (CEST)
> From: Peter Håkanson <peter@ipsec.nu>
> To: faure@kde.org
> Subject: SRV support in konqueror ?
>
> Hi,
>
> I hope i address the right person here, but i would like
> an added feature in konqueror, namely the ability to
> examine and follow DNS SRV records.
>
> SRV records are a new feature in DNS that makes it possible
> to locate "services" in the same way as "MX" records deals
> with mail addresses.
>
> Until SRV records were defined no way existed to tell which
> port a server application used. Thus all servers had to reside
> on the IANA-defined port.
>
> With SRV records the intention is that an application that
> wishes to reach a server should ask DNS for a "SRV" first,
> if it founds one use the server and port given to connect. If
> none is foundm it will ask for "A" ( or usually "ANY" and
> follow "CNAME" or "A" records found.
>
> SRV records would be just perfect for http, the reson is that
> one often wants several "logical" servers in the same machine.
> Previously this was achieved by useing several ip-addresses and
> having (apache) listen on the indivitual addresses. To reduce
> the need for addresses url-based mechanisms was designed,
> but the problems were several.
>
> SRV on the other hand would deal with this in a better way, the general
> format is
> <key> IN  SRV  <prio> <weight> <port> <domain-name-of-server>
>
> So a webserver is described not only which port to be used,
> but also a way of describing several servers, who's order
> to be contacted is definedm and a crude way of load-balancing
> ( broken in my opinion) is defined + the servers domain name.
>
> The key is built up with a special way , assume you want to
> reach www.ipsec.nu , then the key is :
> _http._tcp.www.ipsec.nu
>
> Any SRV record found for that record will describe not only
> which server(s) but also at which ports to use.
>
> Thus a webserver for 100 domains need only one ip-address, but
> has a server on 100 different ports.
>
> Why is this not used today ? Well no browser to my knowledege
> has the ability do do this. Thus no web-owner cares about
> SRV. Someone has to start. I think the KDE project would
> be an excellent starting point ( "the only modern browser" )
>
> I have prepared SRV records for www.ipsec.nu ( and www.ipsec.se)
> for testing purposes ( they point to my ordinary server, but
> that could be adjusted)
>
> Now, if i want SRV records, why don't i grab the source ? Well
> i have i freebsd 4.3 system where i can do this. I am however
> an amatour with c++ and not really confident. I'll try anyway.
> If i get it working, how do i contribute the result ?
>
> Now, is there any interest from the KDE project ? I can
> contribute with DNS knowledge ( this is my living), but i
> am not shure that i can modifu the source as you your
> standards.
>
> Regards

-- 
Andrei Sakharov, Exiled 1980-1986, USSR, http://www.aip.org/history/sakharov/
Dmitry Sklyarov, Detained 2001-????, USA, http://www.elcomsoft.com/

[prev in list] [next in list] [prev in thread] [next in thread]