=?gb2312?B?UkU6ILvYuLSjuiBSRTogd2hhdCBpcyB0aGUgTWF4bGVuZ3RoIG9mIFNJUD8=?=

[Herve Jourdain]

Hi,

 

The fact that it was specified to fallback to TCP if too large for a single
UDP datagram is probably to avoid IP fragmentation¡­

IPv4 has provision for fragmentation, but it costs resources for reassembly,
and thus is not always allowed on routers¡­

The 200 that was chosen in the specs is an arbitrary number that provides
space for responses larger than the request, if I remember correctly, not
the sum of IP and UDP headers (which would be only 28 bytes in IPv4, with
¡°standard¡± headers). I think the figures I gave are for the request, the
response should be able to go up to MTU¡­

 

Regards,

 

Herve

 

  _____  

From: Óê ³Â [mailto:chen.yu26 at yahoo.com.cn] 
Sent: vendredi 25 janvier 2008 07:43
To: Herve Jourdain
Cc: group SIP
Subject: »Ø¸´£º RE: what is the Maxlength of SIP?

 


Hi, Herve,

Thanks for your answer.It help me a lot
<http://us.i1.yimg.com/us.yimg.com/i/mesg/tsmileys2/01.gif> . Well, I am
still not sure about the follow,

 

I know the MTU is a limit unit on the ethernet . So the number you metioned
"200",is it a sum bytes of  the IP and UDP packet  header ?

 

And you tell me when the SIP is too larger for UDP to send ,it will resort
to TCP, However,I think this issue can be solved below the transport layer,
I mean on the network layer. Why not the carried IP packet be divided into
more smaller ones to transmit, something like fragmentation?

 

Nora

  

  _____  

 
<http://cn.mail.yahoo.com/gc/index.html?entry=5&souce=mail_mailletter_taglin
e> ÑÅ»¢ÓÊÏä´«µÝÐÂÄê×£¸££¬¸öÐԺؿ¨ËÍÇ×Åó£¡ 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://sipforum.org/pipermail/discussion/attachments/20080125/2ffaf9bb/attachment-0007.html