DL10 documentation

Noel Chiappa jnc at mercury.lcs.mit.edu
Thu Jan 11 08:47:09 CST 2018

    > I'm to[o] busy right now to dig back through my ancient records (paper
    > and email) to find details

So while I didn't have time to do either of these (my Proteon email, if I
still even have it, will be on a magtape I'd have to get Chuck to read; and
the paper records are mixed in with a giant pile of other stuff - I was on the
IESG while I was at Proteon, and it's all mixed in together), I did take a
quick look online to see if I could locate anything from that time period -
knowing how bad human memory really is, I wanted to make sure my memory wasn't
playing me false.

I didn't have high hopes, since stuff from the late 80's is hard to find
online, and I my expectations weren't disappointed (at least, in the brief
time I could put into it), but I did happen to turn up this:

  John T. Moy, "OSPF: Anatomy of an Internet Routing Protocol"

which I'd vaguely heard about, but don't have (although I have everyone else's
books; I'll have to get a copy), wherein one may find (pg. 303) this:

  "OSPF considered, but did not use, IS-IS as a starting point."

which seems fairly definitive, and straight from the horse's mouth.

I do wish I had access to more contemporary documents to, to give it a bit
more detail. As I recall the circumstances, I had previously wanted to do a
link-state replacement for EGP (to be called FGP) but Dave Clark (who was at
that time on the Proteon board) shot it down (IIRC, in part because he thought
it was too big a job for John - and John was not sanguine either; whereas I
had already seen enough of John to know he was quite capable of it).

That part I remember clearly, but from here on out it gets hazy (I was so busy
with goings-on in the IETF, juggling so many things with that, Proteon, etc),
alas; and it's been too many years since those memories were refreshed by use.
I do recall that we also needed a better IGP, as RIP was not really that good,
and Proteon decided they could do that - and John and I would have agreed that
a link-state design was the only way to go.

It started out as a Proteon-specific thing, for Proteon's customers, but like
SGMP (which started in similar circumstances, before morphing into SNMP), it
soon turned into an 'open' effort, in the IETF. I don't recall how (i.e. why)
that happened, but I assume it was a similar set of reasoning as with
SGMP/SNMP. It might be that if the IETF email archives from that period can be
found, they'd have some useful coverage of that.

My vague memory is that our biggest design influence was the ARPANET work, and
especially the later version which added area support (described in:

  Josh Seeger and Atul Khanna, "Reducing Routing Overhead in a Growing DDN",
    MILCOMM '86, IEEE, 1986

which I have in hardcopy somewhere, which I saw on the top of a pile recently,
so I can scan it if someone's interested), and also the subject of a memorable
briefing to the proto-IETF by Linda Seamonson, which I remember clearly - not
the technical details, alas, just at how good a presentation it was! :-) I
remember in particular they had a very elegant/clever method for defining the
area boundaries.

Like I said, we did 'borrow' some idea from IS-IS, in particular the sequence
number thing - but that may have come direct from Radia's paper:

  Radia Perlman, "Fault-Tolerant Broadcast of Routing Information", Computer
    Networks, Dec. 1983

I don't recall where the concept of a designated router stuff came from, if
IS-IS was any influence there or not.

I did interact with John quite a bit in the very early design stages (I'd been
making a deep study of routing for quite a few years, so I was really the only
person there who was steeped in routing he could talk to), but as the work
prgressed - particularly once it moved to the IETF - I got out of the loop, as
I was too busy with other things, and he clearly had things in hand. I also
seem to vaguely recall disagreeing with him about some design points, but I
can't remember what.

Anyway, probably the wrong list for this. (Internet-history would have been
better.) Sorry, I didn't mean to get into a long thing, thought I was just
correcting a bit of nth-hand 'telephone-game' type garbling of a minor point.>


More information about the cctalk mailing list