header{ Saku Ytti (2013-12-01) () Subject: Fwd: Ethernet history, why 1500B From: Saku Ytti To: charlie@xxx.com Date: Sun, 01 Dec 2013 20:05:59 +0200 header} body{ part{ ID: 1, Content-type: multipart/alternative part{ ID: 2, Content-type: text/plain ---------- Forwarded message ---------- From: Bob Metcalfe Date: 1 December 2013 16:24 Subject: Re: Ethernet history, why 1500B To: Saku Ytti There is a limit on Ethernet packet sizes to enable sharing on shared cable back when, reduce latency, provide for packet buffers, specify error control rates. PARC Ethernet max size was set big enough to carry a whole disk sector with metadata. That's all I know. Typt withums, corected buy iFoan. > On Dec 1, 2013, at 7:29 AM, "Saku Ytti" wrote: > > Hi Bob, > > Googling and literature didn't surface with conclusive reason why ethernet originally came up with 1500B frame limit. > If you can provide me with ISBN or URL for authorative answer, it would be very nice. > > Some say, to avoid collision with ethertypes with EthernetII but that seems like causality violation, how could have you known about those requirements prior to them happening. > > Other, more plausible is, that mathematical guarantees of 4B FCS change post 1500B, so larger frames would have needed larger FCS field, which was deemed as too high overhead. > > > Thanks, > -- > ++ytti -- ++ytti part} part{ ID: 3, Content-type: text/html Non-text part: text/html part} part} body} message} header{ Saku Ytti (2013-12-01) () Subject: Fwd: Ethernet history, why 1500B From: Saku Ytti To: charlie@xxx.com Date: Sun, 01 Dec 2013 20:06:09 +0200 header} body{ part{ ID: 1, Content-type: multipart/alternative part{ ID: 2, Content-type: text/plain ---------- Forwarded message ---------- From: Rich Seifert Date: 1 December 2013 18:37 Subject: Re: Ethernet history, why 1500B To: Saku Ytti On Dec 1, 2013, at 6:55 AM, Saku Ytti wrote: > Hello, > > I'd like also to pick your brain, how you recall the history for 1500B maximum ethernet frame size. > > Apologies but, I've not read your Switching book, friend of mine recalled it may have discussed this question. I found this quote which may be from you in http://www.tomshardware.co.uk/forum/19754-42-maximum-size-ethernet-frame > --- > -A longer maximum frame increases the memory requirement for a NIC using > a simple, fixed buffer design. This is the *real* reason for the 1500 > byte limit; at the time we designed it (1979), buffer memory was much > more expensive than it is now, and DMA controllers were too complex to > be implemented in anything less than a full-custom chip. > ---- > > > I guess it also might be, that there was no single deciding moment when 1500B was picked and may have different motivation to different people. > The quote is correct; the primary reason for the 1500 byte frame length limit was to reduce the buffer memory requirement for network interfaces. The intent was to allow for 1K byte of data, plus 500 bytes for protocol overhead, encapsulation, etc. There *was* a single, deciding moment during the DEC-Xerox discussions when we locked down the length. In retrospect, a longer maximum might have been better, but if it increased the cost of NICs during the early days, it may have prevented the widespread acceptance of Ethernet, so I'm not really concerned. -- Rich Seifert Networks and Communications Consulting rich@richseifert.com 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 -- ++ytti part} part{ ID: 3, Content-type: text/html Non-text part: text/html part} part} body} message}