[ih] IEN's as txt
jnc at mercury.lcs.mit.edu
Fri Feb 10 05:17:18 PST 2017
>> Same goes for ICMP: it was a late arrival and the rationale for
>> abandoning the earlier approach is not entirely clear.
> Before ICMP, the only protocol for exchanging information between hosts
> and routers about the state of the network, was GGP, IEN-109.
> it was obvious to us at that point that the routing was going to need a
> lot of work
Oh, you might be asking yourself 'surely it was obvious early on that you all
needed something better, why didn't it get done sooner'. Two answers: manpower,
First, the effort was pretty small (on the order of a few dozen people), and
most time and energy was being spent in simply writing code for the many
different OS's (there were a lot more back then, and many had to be coded in
assembler). That whole effort was made more difficult because OS's of that era
were generally not 'networking friendly', and often needed major internal
surgery to successfully support any networking. So there just weren't people
sitting around twiddling their thumbs waiting to think about, e.g. routing. We
all had 'day jobs' (I was building routers), and we had to fit this redesign
work in in what little time we could spare from actually getting the thing
running (which was absolutely flat-out, more-than-full-time already).
Second, we had so little practical experience to go on at the time, and you
shouldn't underestimate how much of a handicap that was. The picture of what
is needed is much clearer today than it was at the time. E.g. subnetting only
got added as LANs appeared; when the early work was done, there were only a
few large long-haul networks (hence the original addressing structure in IP).
We just didn't know.
More information about the Internet-history