[ih] This Review is for Everyone

[Karl Auerbach]

Yeah, that Netbios-over-TCP/UDP was a tough thing.

(By-the-way, on my desk I still have a post-it-note holder with the 
Ungermann-Bass name that you gave me when we were working on - or 
fighting over - all of this stuff.)

I'm not particularly proud of the standard we developed but then again 
we were trying to stuff a local broadcast domain protocol into a 
potentially wide area box.  We kinda ended up doing a Procrustean effort 
of stretching this and cutting off that.  I think the result was 
preordained to be a battered and bruised shadow of the original Netbios.

I am sure that many who have extended our work to be the basis of 
Microsoft's CFS have cursed my name for some of the choices we made.  I 
kinda deserve it.

I can't remember whether the Air Force ULANA project had started by that 
time, but I had long known many of the developers of the original 
Netbios code at Sytek - David Kaufman and Greg Ennis.

Much of that original code was quite clever in how it used the 
particular aspects of the Sytek and IBM ring networks and much of what 
became the original IBM/Sytek Netbios was designed to fit that hardware 
- so the code was actually quite small.  (As an aside, the Sytek ring 
contained various ideas from Dave Farber's DCS via Frank Heinrich who 
flew through many of these efforts like a bee pollinizing flowers with 
ideas and insights.)

I do remember that as we tried to move Netbios into an Internet world we 
ran into a lot of technical problems, not the least of which is that the 
Netbios service had a lot of assumptions about broadcasts and fast 
response from all pieces of a Netbios community - something that is 
sensible on a LAN, particularly a LAN based on a ring with a rotating 
token, but doesn't fit well on a wide area (and thus slower and less 
reliable) network.

Amatzia Ben-Artzi (Sytek) and I came up with what was a unified design 
that tried to preserve much of the Netbios service interface in both 
local and wide area contexts.  There were other designs that bifurcated 
things into a local mode and wide area mode.  The latter was adopted.  
That, in hindsight, was probably the wiser choice, but it certainly made 
things more complicated.

The reason that I mentioned ULANA was that netbios-over-TCP was a 
requirement.  So we had to implement what we had done.  I remember 
working with a small company in Petaluma, California that had expertise 
in putting original netbios into various devices.  (I wonder how many 
times I crossed the Golden Gate bridge on my often daily travels between 
Santa Cruz and Petaluma?)

Although our team (TRW) won the ULANA bid, the project never took flight 
because another major bidder, AT&T, protested and the whole thing sank, 
so we never really had the chance to test whether our ideas really 
worked.  Microsoft picked up the flag and ran with it via SMB and CIFS - 
and as I mentioned, my name is probably mudd in those quarters.

But I want to put in a plug for the vast good that the often forgotten 
ULANA procurement incentivized.  ULANA built a serious and hot fire 
under commercial-off-the-shelf products for TCP/IP based networks.  I 
remember Dave Kaufman telling me about how he visited Cisco when it was 
still essentially a garage operation.  I think my initial implementation 
of John Romkey's Packet Driver module for PC-DOS was picked up by Russel 
Nelson (Crynwr) created a standard and easy to use packet send/receive 
API for code on PCs (until Windows came along.)  That, in turned enabled 
a lot of applications to be developed with the need for those to worry 
about device drivers for the rather large number of Ethernet cards that 
were appearing.

     --karl--


On 3/22/26 12:12 PM, Dave Crocker via Internet-history wrote:
>    A timely case in point was that we were one of 3 companies that had 
> done independent development of IBM's Netbios over TCP/IP.  IBM (and 
> Sytek) had only published the API, with no details about the 
> underlying protocols.  So all 3 of us made whatever choices we felt 
> appropriate.  For example, we even added the ability to link Netbios 
> across different locations.
>
>    Some of had experience with the 'co-opetition' model of 
> collaborative specification effort common the the networking 
> community.  Others did not.  This eventually led to an impasse that 
> required call on Vint to come in and mediate.  RFC 1001/1002 was the 
> result:  March, 1987.
>