[ih] A revolution in Internet point-of-view - Was Re: Internet analyses (Was Re: IPv8...)

[Dave Crocker]

On 4/29/2026 6:13 AM, Andrew Sullivan via Internet-history wrote:
> Andrew Russell's argument in _Open Standards and the Digital Age: 
> History, Ideology, and Networks_ (Cambridge UP, 2014) casts this 
> cultural difference in another light: that the OSI approach embodied 
> (or, and I think this is my gloss, "was hobbled by") old-fashioned 
> governance models that were in an important sense more open than the 
> ARPANET/Internet work that led to TCP/IP.  For, he argues, the 
> Internet innovators had to satisfy only one customer: the US Defense 
> Department (or, depending on exactly when, [D]ARPA).  By constrast, 
> the OSI work as well as work that happened before that were 
> international efforts that involved many different parties, sometimes 
> with competing interests, and that used the mechanisms of co-operation 
> used in other places in societies too (like voting etc.) to make 
> decisions.  In this account, the problem OSI and other such efforts 
> faced is basically that they could not move as fast as those who 
> needed only to satisfy a single "customer" of the work. 


For the Lynch/Rose book, I did a chapter on the standards process and 
included a superficial comparison with the OSI processes:

    *Evolving the system*. In Internet System Handbook, D. Lynch, and M.
    Rose, eds. Addison-Wesley, Reading, Mass. 1993
    https://archive.org/details/internetsystemha00erne

I did a version of this for ACM Standardsview:

    *Making Standards the IETF Way*
    https://dl.acm.org/doi/epdf/10.1145/174683.174689

Pete Resnick did an RFC that included a point about the IETF I consider 
fundamental and significantly underappreciated:

    *On Consensus and Humming in the IETF*
    https://www.rfc-editor.org/rfc/rfc7282.html


Russel's focus 'governance model' is helpful, but it it is worth 
distinguishing between at least two venues for this.  Also, I think he 
got a critical point wrong, albeit for an understandable reason.

  * Governance applies to the standards process itself, but also to the
    operational model.
  * Having centralized funding does not automatically mean centralized
    decision-making


*Operational Model*

    The original Arpanet/Internet model had a centralized backbone, with
    highly independent users (enterprises/hosts).  Some ad hoc Internet
    backbones emerged in the early 1980s, but NSFNet solidified the
    requirement to permit multiple, independent backbones and finally
    produced a routing protocol to support this.

    By contrast the OSI world was dominated by a model of tightly
    controlled, monopolistic carriers providing backbone services. On
    the surface, this might be thought to be similar to the Internet
    model that developed, but the relationship among monopoly carriers
    was/is profoundly different from the very loose 'co-opetition' model
    prevalent in the Internet, including for backbone services.

    Note that X.400 addresses required specifying the email backbone
    carrier.  This led to bizarre business cards that might contain a
    number of (very long) separate X.400 addresses, differing only by
    that one attribute field.

    *
    *

*Standards Process*

    While it's true the original Arpanet and Internet work derived from
    central funding and oversight -- and the oversight did, sometimes,
    get asserted at important decision junctures -- the daily reality
    for making decisions involved gaining agreement among a collection
    of highly independent participants. (I see that Steve has made a
    similar point.)

        It is worth having a thoughtful discussion to consider whether
        the historical participation and decision model still, fully
        applies in the IETF, but of course, that's not a history topic...

    The OSI world's model for decision-making is unanimity.  The
    Internet's is 'rough consensus', which translates to 'strongly
    dominant agreement'.  Leaving room for routing around minority
    dissent has a profound effect on the the ability to make forward
    progress. However, this has not meant ignoring minority concerns,
    which is nicely discussed in Resnick's RFC. (However it also seems
    to be far less applicable now.)

    The Internet approach was aided by an additional, major difference: 
    urgency.  The early Arpanet/Internet work was dominated by a desire
    to have something fielded quickly. Yesterday would have been ideal. 
    By contrast, the OSI world generally had a very casual approach to
    schedule, measured in years rather than months.  (Alas, this is now
    true for most Internet work, too.)

    The Internet's reliance on email venues for work, with physical
    meetings primarily used to focus on essential points of difficulty,
    permitted steady, incremental progress.  For the OSI world, all work
    took place at occasional, long, face-to-face meetings. 
    Surprisingly, for the Fax-over-email work, I witnessed the latter
    try to spontaneously make major design changes, literally overnight,
    intending to make a final decision the next day.

    It is also significant that the Internet specifications documents
    were openly available, all through the development process and
    beyond. The OSI documents were paywalled.  Added to this was that
    participation in the Internet processes tended to be relatively open
    -- and from around 1987 completely open -- whereas OSI participation
    was tightly constrained.  This meant the Internet work got much more
    community review than the OSI work did.  (A common experience for
    folk involved in the early Internet work, when traveling around the
    world, was to come across random technical folk who had read the
    RFCs.  Mine in 1997 was on the northern end of Borneo...)

    Lastly, the IETF world has tended to develop proof-of-concept
    prototypes, as specifications are developed, and sometimes has even
    mandated this.  My understanding is that the OSI world did not.

    I've characterized the resulting difference between these two
    processes as:  The OSI world took the union of everyone's wish list;
    the Internet world took the intersection.  The former tended to be
    lean and pragmatic.  The latter tended to be large, cumbersome and
    operationally challenging.

d/

-- 
Dave Crocker

dhc at dcrocker.net
bluesky: @dcrocker.bsky.social
mast: @dcrocker at mastodon.social
+1.408.329.0791

Volunteer, Silicon Valley Chapter
Northern California Coastal Region
Information & Planning Coordinator
American Red Cross
dave.crocker2 at redcross.org