[ih] Early FTP development in the ARPAnet

Bob Braden braden at ISI.EDU
Sat Sep 7 21:04:10 PDT 2002

I have not received or seen Neal's questions (Chris, what's going on?)
but I do have some answers from the firing line.

  *> >
  *> >
  *> >
  *> >1.)  For those involved in the development of FTP through the years=20
  *> >can you recall what influenced you to work on FTP?

My boss told me that UCLA's IBM 360/91 supercomputer, for whose
programming I was responsible, had to become a server on the ARPAnet.
You can't provide service if you can't move files.  And if you leave
the file moving protocols to the mercy of the folks in Cambridge, MA,
it is NOT going to support an IBM mainframe.  So I had to get involved.

Besides, it was a new and fun problem.

  *> >
  *> >
  *> >2.)  In the beginning of the Arpanet, what was the demand for=20
  *> >something like FTP?  Inside the group of IMP locations, what was=20
  *> >there a level of excitement/ambivalence over FTP?  How long did it=20
  *> >take for people to begin understanding the benefits of sharing=20
  *> >information, programs, or research?

Neither excitement or ambivalence.  The need to transfer files was
obvious.  The benefits of resource sharing were by definition,
because that is what ARPA thought the network was for.

  *> >
  *> >
  *> >3.)  How did ICCC affect the progress made on FTP?

Oh, my.  Larry Roberts came to a Network Working Group meeting and gave
us designers/implementers holy hell!!  He made it perfectly clear that
if we expected to keep ARPA funding (he did not of course say that in
so many words, but we didn't need pictures), we WOULD make ICCC a
shining success, and we had better get a move on!!  Actually, Telnet
was more critical and difficult than FTP.
  *> >
  *> >Was there a change in atmosphere or a since of urgency?

Yes. See above.

  *> >
  *> >Did more people contribute as a result?


  *> >
  *> >4.)  At the ICCC in October 1972, how important was the concept of=20
  *> >FTP to the people organizing the event and the people attending?

It was one of the basic tools.  See the ICCC manual for details.

  *> >
  *> >Did people immediately realize its usefulness or was everything out=20
  *> >shadowed by the concept of packet-switching?
  *> >

Dumb apple/orchard question.

  *> >
  *> >5.)  According to the ICCC manual many of the scenarios were=20
  *> >actually contributed from former programs used at workshops at MIT.=20
  *> >How early did scenarios appear involving FTP or software that=20
  *> >exploited FTP?

I think that would have applied only to the MIT demos, but my memory
is hazy.  Certainly the UCLA demo had nothing to do with a workshop
at MIT.

  *> >
  *> >
  *> >6.)  In between the release of each RFC what was process that would=20
  *> >precede the release of the next RFC?
Do you mean releases of the FTP spec?  There is an extensive record of
this in the RFC series... various meetings, minutes of meetings, plans
for future meetings, etc., etc.

  *> >
  *> >=20          a.)  How were decisions made?
  *> "consensus"  (and a lot of very loud debate).  Sometimes the decision=20
  *> was based on a good punch line.

I agree with John Day here, except that sometimes (often?) the final
decision many technical details were actually made by the person who
wrote up the meeting notes or who next revised the protocol draft
;-)  Sometimes I was surprised, often bemused, to learn after the
fact what we had "decided".  But this was OK because the people
involved were very smart, and unless they really screwed up we
were content to accept their refereeing and get on with it.

  *> >
  *> >=20          b.)  What meetings do you recall?  What was the mood of=20
  *> >those meetings?
  *> Meetings were lively.

Again I agree with John.  At this distance, I have only hazy recollections
of most of the meetings, and depend mostly on the written record.

No, I recall an very early meeting when magisterial Steve Crocker swept
in on his flying carpet and gave one of his now-familiar
back-to-fundamentals off-the cuff lectures, this one on what a file
transfer protocol ought to do.  He explained he had written the notes
on the back of an envelope during the train ride down (so help me!)
Completely mathematical and abstract, of course.  But I was impressed
and intrigued, in spite of the unexpected dunk into the cold water tank
of Computer Science, when we were deeply engaged in heavy-duty

  *> >
  *> >=20          c.)  How was everything organized?
  *> Around a table.
  *> >
  *> >=20          d.)  Who did the talking?
  *> Everyone.  Transmission was distinctly full duplex, broadcast.
  *> >
  *> >=20          e.)  What people stand out in your mind?

Various people took the lead on the ARPAnet FTP spec at various times:
Abhay Bushan at MIT, then later Nancy Neigus at BBN, then Alex McKenzie
at BBN.  I recall Alex as contributing a lot of protocol maturity and
some really clever ideas, like the restart mechanism.  John Day brought
in a late proposal for a file access protocol, which seemed like a very
nice idea but terribly hard to implement with any generality in some
OSs (like mine), so I lobbied against it (sorry, John).

Steve Crocker (see above).  And of course the BBN guys were always
doing first class work on the Tenex platform, and the rest of us
scrambled to keep up.

If I thought more than 5 minutes, I would certainly think of others.
My apologies to good friends from that era whom I have slighted.

Bob Braden

More information about the Internet-history mailing list