[e2e] Protocols breaking the end-to-end argument
jnc at mercury.lcs.mit.edu
Sat Oct 24 20:20:10 PDT 2009
Wow, you all have been busy. I'll have to get to some tomorrow, but this one
> From: John Day <day at std.com>
>> The reference to the "ISO protocol wars" is completely mystifying, as
>> the architecture of the ISO stack (at least, the CLNP/TP4 flavour,
>> which was the subset which gave TCP/IP the best 'run for their money')
>> is basically dentical to that of TCP/IP
> Cmon Noel, you know better than that. That was never what the protocol
> wars were about.
>From where you were sitting, perhaps. My view was from a somewhat different
locations... :-) Seriously, though, the only protocol war I saw involving ISO
was the TCP/IP <-> ISO one - which is covered at some length in Abbabte's
excellent book, IIRC.
> It was not a war between CLNP/TP4 and TCP/IP, but a war between
> (CLNP/TP4; TCP/IP) and X.25.
Yes, but even at the time, to some of us that one had all the suspense of the
First Zulu War - gee, which is better, spears or repeating rifles? Doesn't
take a surfeit of brain cells to call that one..
> The argument at the time by the PTTs was that a Transport Protocol was
> unnecessary. Our argument, of course, was that it was absolutely
> necessary. This was the big argument from about 1976 to 1985. This is
> primarily what the end-to-end paper discusses
Yeah, but the reason TCP/IP clobbered X.25 had not much, I feel pretty sure,
to do with the inherent fundamentals of virtual circuits versus datagrams.
To me, the whole X.25 world was badly crippled by the whole X.75 (I think it
was X.75 - you can correct me if I've mis-remembered) mess that you had to use
if you wanted to set up a VC that spanned a number of providers. TCP/IP and
CLNP/TP4 (hereinafter, 'The Datagrams') never had to deal with that - in large
part because they were designed from the get-go to span multiple providers as
easily as they spanned a single one. My perception is that X.75 happened, in
large part, because it was an add-on to a legacy, X.25, which was designed for
single-provider networks. Of course, one could say that that was inherent,
because 'of course' to span multiple provider networks you'd have to a
mechanism to connect VC's together, but I'm not so sure; it should have been
possible to have a single unified VC 'network' that spanned multiple providers
- but of course the internal protocols (for acknowledgement, etc) would have
had to have been standardized, so they could operate on an inter-provider
But to me even that wasn't even the biggest handicap for X.25. The Datagrams
were set up pretty much from the get-go to work across a range of
technologies, including local area networks - and I claim it was the
Datagrams' friendliness to LANs which was the real 'killer app' in the X.25
<-> Datagram war. Of course, one could say that again, this was 'inherent' in
the architectures - but again, I'm not so sure; had a VC protocol been
designed _from the start_ to work well across a variety of media, including
LANs, it would probably have been considerably more LAN-friendly. It was the
'boat anchor' of the pre-LAN design of X.25, which the PTTs/etc couldn't or
wouldn't discard (because of installed base, existing software, and brain,
code, etc), that was the real killer of the X.25 dinosaurs in a world of LAN
So I don't think X.25 vs The Datagrams was really a 'fair' test of the
fundamental pros and cons of VCs and datagrams: X.25 versus The Datagrams was
more like a rusty, bent, dull assegai versus the latest AR-15 - hardly a fair
test of the fundamentals.
> tries to create a "higher moral ground" by creating a more general
> (and hence more fundamental) principle to base the debate on.
Maybe my brain is fading, but my memory is that the trio were more interested
in a design philosophy framework for designing protocols in a basically
datagram environment; bringing more firepower to the VC/datagram debate was
not at all of much interest.
Again, to us, at the time, VC's had all the interest of rotationally
optimizing assemblers - clearly a technology which the march of improvement
> It was only later that the unwashed in the IETF turned it into a CLNP
> vs IP war.
Read Abbate; TCP/IP versus CLNP/TP4 was a real set-to. As a TCP/IP backer, I
was far more worried about CLNP/TP4 than I ever was about X.25, which was
clearly a rusty assegai in a world of repeating rifles.
Unwashedly yours, :-)
More information about the end2end-interest