Does anybody know how well ROHC would perform in the specific case Steve's draft covers? If it does as well, or almost as well, then I would agree with Pekka- any case where the bandwidth is critical should be using ROHC anyway.
Brian Pekka Savola wrote: > > On Tue, 4 Dec 2001, Hesham Soliman (ERA) wrote: > > > Saving 16 or 32 bytes in the datagram is IMO basically irrelevant. > > > > => "irrelevant" to what ?? Not to BW consumption obviously. > > IMO, there comes a point when adding complexity to save a few bytes off > every datagram may not be worth worrying _too much_ about it. > > On slow, high-delay links (e.g. 9600bit/s) this might make some sense, > but these need a header compression _anyway_ so I don't think there are > any additional gains there. > > An example. While typing this, I captured with tcpdump the IPv6 datagrams > I was typing. It appears that in a second, on average ~5-15 packets were > sent in one direction. Each of these had an IPv6 header of 40 bytes and > TCP header plus basically a few characters -- ~60 bytes. Taking the > average of 10 packets/sec, this amounts to about 4800 bit/sec of frantic > typing. With additional 32 bytes, the requirement would have been 7360 > bits/sec. > > Contrast that to e.g. 64kbit/s ISDN connection, 56kbit/s modem, 10Mbit/s > Ethernet and you don't see a huge difference; if you want to transfer > data, you'd use bigger packet size anyway. > > Basically the required additional capacity would be something from > (40+32+1)/(40+1) = 78% (very small packets) to (40+32+1428)/(40+1428) = 2% > [or even less]. > > The first is not a problem if you don't have too slow link (where you'd > need ROHC or something anyway), and latter is rather irrelevant IMO. > > Not every byte counts. > > Which is why I don't see much of a point in adding complexity to only sate > one, rather minor need (slow, high-delay links). > > But there are other gains to be gained in the address deletion though, > which make it more interesting than just saving a few bytes. > > -- > Pekka Savola "Tell me of difficulties surmounted, > Netcore Oy not those you stumble over and fall" > Systems. Networks. Security. -- Robert Jordan: A Crown of Swords > -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
