Hi,

There has been a lot of talk in this thread about DNS. I want a solution that works without DNS. If one uses SIP or RTSP to establish a media pat using DCCP. In most cases one do include the explicit IP address rather than the FQDN. Thus no DNS queries etc. I think something that makes this work on the same port is desirable.

Cheers

Magnus

Phelan, Tom skrev:
Hi Dan,

Thanks for the detail in this and your previous message.  I agree that
this is worthwhile to explore for the next version of dccp-natencap and
it'll give me a good head start on working things out.

Tom P.

-----Original Message-----
From: Dan Wing [mailto:[EMAIL PROTECTED]
Sent: Saturday, February 23, 2008 1:03 PM
To: [EMAIL PROTECTED]
Cc: Phelan, Tom; 'Colin Perkins'; ''dccp' working group'
Subject: RE: [dccp] DCCP-over-UDP [was
draft-phelan-dccp-natencap-00.txt]
Can you explain why 100ms seems reasonable?
Longer than that would become noticable in those cases where DCCP-RAW
was
being blocked.

I commonly use paths that have much longer delays than this,
and would
have expected native DCCP/IP encapsulation to still be used in these
cases, rather than defaulting to an encapsulated format.
The algorithm I am proposing would still achieve that.  The algorithm
I am
proposing has you send a DCCP-RAW request and wait 100ms for a
response.
If
no DCCP-RAW response is received in that 100ms, you then try DCCP-UDP.
You do
_not_ abandon the DCCP-RAW connection attempt, however.  This will
provide
you
with a DCCP-RAW connection even over >100ms round-trip delay paths.  I
agree
that if the first DCCP-RAW request is dropped (due to network
congestion
or
other reasons) you may, for that connection, end up with DCCP-UDP
(assuming
the DCCP-UDP request and its response are not dropped).  I don't see a
solution to that problem, other than waiting a really long time for
the
DCCP-RAW to fail.  Waiting a really long time for DCCP-RAW to fail
will
make
DCCP-UDP seem very slow at connection establishment, and I will cause
DCCP-UDP
connection setup to be perceived as 'slow' by users and user
applications.


The case with no middlebox (NAT or firewall):



     DCCP client                     DNS server    DCCP server
     -----------                     ----------    -----------
          |                               |              |
          |---- DNS SRV?----------------->|              |
          |---- DNS A?------------------->|              |
          |<--- DNS SRV ------------------|              |
(not used)|<--- DNS A - ------------------|              |
          |----------------------- DCCP-RAW request----->|
     (wait 100ms)
          |----------------------- DCCP-UDP request ---->|
          |<--- DCCP-RAW response -----------------------|
     (continue with DCCP-RAW connection)
          |<--- DCCP-UDP response -----------------------|
     (this DCCP-UDP response isn't needed, so abort it)
          |----------------------- DCCP-UDP close ------>|
          |                               |              |

with a middlebox:

     DCCP client      middlebox    DNS server    DCCP server
     -----------      ---------    ----------    -----------
          |               |               |              |
          |---- DNS SRV?----------------->|              |
          |---- DNS A?------------------->|              |
          |<--- DNS SRV ------------------|              |
(not used)|<--- DNS A - ------------------|              |
          |---------------X               |              |
     (wait 100ms)         |               |              |
          |----------------------- DCCP-UDP request ---->|
          |<--- DCCP-UDP response -----------------------|
     (continue with DCCP-UDP connection)  |              |
          |               |               |              |

-d




--

Magnus Westerlund

IETF Transport Area Director & TSVWG Chair
----------------------------------------------------------------------
Multimedia Technologies, Ericsson Research EAB/TVM
----------------------------------------------------------------------
Ericsson AB                | Phone +46 8 4048287
Torshamsgatan 23           | Fax   +46 8 7575550
S-164 80 Stockholm, Sweden | mailto: [EMAIL PROTECTED]
----------------------------------------------------------------------

Reply via email to