As Brian says, there are several different answers.

In reality, the max length IP packet supported is constrained by the MTU (max transmission unit) supported by the host at each end (usually a function of receive buffers) and of the subnets in the path.

Since IP is an internetworking protocol, it runs over a number of networking technologies and protocols, each with its own frame or packet size limitation. For example, running IP over an X.25 network (e.g., ISDN D-channel X.25) might limit the MTU size to 128 octets. Ethernet allows 1500 bytes.

RFC1191 contains a table of MTU sizes for various subnet technologies and the RFC numbers in which they are defined. I copy the table below (one advantage of having ASCII specs :-).

Plateau MTU Comments Reference ------ --- -------- --------- 65535 Official maximum MTU RFC 791 65535 Hyperchannel RFC 1044 65535 32000 Just in case 17914 16Mb IBM Token Ring ref. [6] 17914 8166 IEEE 802.4 RFC 1042 8166 4464 IEEE 802.5 (4Mb max) RFC 1042 4352 FDDI (Revised) RFC 1188 4352 (1%) 2048 Wideband Network RFC 907 2002 IEEE 802.5 (4Mb recommended) RFC 1042 2002 (2%) 1536 Exp. Ethernet Nets RFC 895 1500 Ethernet Networks RFC 894 1500 Point-to-Point (default) RFC 1134 1492 IEEE 802.3 RFC 1042 1492 (3%) 1006 SLIP RFC 1055 1006 ARPANET BBN 1822 1006 576 X.25 Networks RFC 877 544 DEC IP Portal ref. [10] 512 NETBIOS RFC 1088 508 IEEE 802/Source-Rt Bridge RFC 1042 508 ARCNET RFC 1051 508 (13%) 296 Point-to-Point (low delay) RFC 1144 296 68 Official minimum MTU RFC 791

Table 7-1: Common MTUs in the Internet

In a general case implementation, I would support at least 1500 bytes MTU sizes. As for your specific application (video), it might not require 1500 byte packets. Supporting some type of Path MTU Discovery (RFC1191 or RFC1981) might be useful to avoid fragmentation since some paths don't support 1500 byte MTUs. For example, many tunneling technologies (GRE, L2TP, IPSEC) reduce the effective packet size.

Chip

At 10:09 PM 10/25/00 -0400, Rosen, Brian wrote:

There is the "real" answer, and the practical answer.

The real answer is that depending on what kind of network you are on, you could see, in theory, packets of up to 64K bytes. 4K bytes can actually be seen in some Gig-E systems. Whether you will find a video codec using such large packets sizes or not is uncertain; I've never seen one.

It used to be that we thought 1510-odd bytes was the biggest practical packet you see on a LAN, but that's no longer the case. There is a lot of code around that fails above that 1500 ish number.

Brian

...

-----Original Message----- From: Chris Farmer [mailto:cfarmer@smithmicro.com] Sent: Wednesday, October 25, 2000 9:05 PM To: ITU-SG16@mailbag.cps.intel.com Subject: Sanity check on Packet Sizes....

Okay,

I was researching an arcane fact on the theoretical and practical sizes for RTP packet sizes for video conferencing.

What I have come up with is: There is specific RFC's about implimenting H.263, H.261, G.723, but I'm looking for just an overall practical size to check input buffer sizes.

RFC 1889 RTP packet structure referes to 768, RFC 791 RFC 768 UDP refers to obsolete RFC 760. RFC 791 states in the Glossory, ARPANET messages max size is about 1012 octets (8096)bits. ARPANET packet max size is about 126 octets (1008) bits.

As a client side EP, should I be concerned about the ARPANET message size or the ARPANET packet size?

RFC 760 states: 65,536 octets theoretically, put not practical Page 22, every destination should be able to receive a datagram of 576 octets in one piece or as fragments to be reassembled

so my guess here is 4608 bits max? Or is this too low level and there is another RFC that I should reference?

Any help in this direction?

Thanks,

Christopher B. Farmer Software Engineer Smith Micro Software, Inc. (503) 641-1221 ext 28 ====================================================================== Daily C\C++ Tidbits at www.EngageUs.com\c "Challenges in life are inevitable. Failure is optional" - Roger Crawford ======================================================================

For help on this mail list, send "HELP ITU-SG16" in a message to
listserv@mailbag.intel.com

For help on this mail list, send "HELP ITU-SG16" in a message to
listserv@mailbag.intel.com

------------------------------------------------------------------- Chip Sharp Consulting Engineering Cisco Systems "The best way to pick a lock is with a key." http://www.netaid.org -------------------------------------------------------------------

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For help on this mail list, send "HELP ITU-SG16" in a message to listserv@mailbag.intel.com