[ih] when did APRANET -TIPs become known as -TACs
Karl Auerbach
karl at iwl.com
Sun Sep 28 17:28:05 PDT 2025
It annoyed me that IPv6 did not adopt the Fletcher checksum that was
used in ISO/OSI TP[1234] and CLNP. (Or perhaps even a Hamming error
correcting code.)
(At first glance the Fletcher checksum looks like a computational burden
with expensive machine operations, such as integer multiplies, but there
are both full and incremental algorithms for it that are quite
efficient, avoid expensive machine instructions, and fast. And the
Fletcher checksum does catch byte big-endian/little-endian reversals, or
what was called in the early days of Unix the "nUxi" [pairs of bytes
reversed] problem.)
We now live in a world where jumbograms - and large path MTU values, on
the order of 8K+, are rather common. And those larger packets are more
likely to get hit with a burst of line noise than were the shorter
packets of days gone by. But that revives the old debate whether to do
error checks/correction on a per-link basis or end-to-end basis (in
addition to per-link checks/correction.) (Those of us who were burned
by the old Sun S-bus has no parity check, coupled with Suns NFS using
zero as null UDP checksum, kinda cling to the end-to-end check side of
the argument. [When I was at Sun a critical machine had an
S-Bus/Ethernet path that became noisy and the effect caused the major
source code repository used for all Sun code to become corrupted. It
was a mess to clean up.])
--karl--
On 9/28/25 4:26 PM, John Day via Internet-history wrote:
> No one took those seriously.
> TP0 was for CCITT SGVIII, TP1 was for CCITT SGVII, TP2 was for the Brits who had to use X.25, and TP3 was for the Germans.
>
> No one paid them any mind. The real focus was on TP4 which was a major advance because it adopted Watson’s insight on synchronization which made it much simpler and more secure.
>
> Take care,
> John
>
>> On Sep 28, 2025, at 19:21, Dave Crocker <dhc at dcrocker.net> wrote:
>>
>> On 9/28/2025 4:13 PM, John Day via Internet-history wrote:
>>> The OSI vision was, if anything, Ethernet (an ISO standard), CLNP, TP4, and ACSE.
>> That sounds reasonable, except for TP0, TP1, TP2, and TP3.
>>
>>
>>
>> and the various CONS alternatives to CLNP.
>>
>> OSI went for the union of everybody's wish lista. The Internet went for the intersection.
>>
>> d/
>>
>> --
>> Dave Crocker
>>
>> Brandenburg InternetWorking
>> bbiw.net
>> bluesky: @dcrocker.bsky.social
>> mast: @dcrocker at mastodon.social
More information about the Internet-history
mailing list