Why Protocols Fail (was Re: [e2e] Jon Crowcroft's thought questions...)

Simon Spero ses at unc.edu
Mon Aug 4 06:01:24 PDT 2003

--On Saturday, August 2, 2003 9:34 AM +0100 Lloyd Wood 
<l.wood at eim.surrey.ac.uk> wrote:

> but a rigid binary format and a lack of flexibility was a factor in
> helping doom Gopher long-term.
> (Okay, I'm ignoring the whole short-sighted US university's licensing
> terms protecting their IPR and killing adoption of their pet protocol
> here.
> http://www.wikipedia.org/wiki/Gopher_protocol  )

Gopher isn't really a good example, since the financial factor you mention 
was so totally dominant.  Because just about every client that could use 
the gopher protocol could also use other protocols, as soon as UMN started 
trying to 'develop'* gopher, people shifted protocols without a backwards 
glance. It's really a great example of interoperability preventing lock-in.

The reason that www and http took root is that Tim Berners-Lee worked so 
hard to keep things open, giving a base that people could build on without 
being tied in to one protocol or one company.

The Internet between 1990-1992 was like Europe in 1914;  AUPs becoming less 
restrictive, new backbone providers (ANS - why did they always leave out 
the U?); new graphical clients designed for non-experts (WAIS); net wide 
indexing (Archie). The net was a powder keg waiting for a spark. Something 
was going to take happen, but the precise form that the explosion took was 
always going to be based on contingent factors.

This whole thread suggests some interesting areas of research;  it could be 
very productive to take a look at several examples of protocols that didn't 
take become widely used, and see if there are any common threads.  There 
are obvious issues, such as  complexity, lack of resources/commitment for 
development, duplication of effort, poor specifications, lack of 
functionality, or lack of need.

What about the protocols that survived, despite having one or more of these 
problems; what were the over-riding factors.

Has anyone published on this?



* to develop: (v) an academic term meaning "to destroy promising technology 
through the careful application of neckties"

More information about the end2end-interest mailing list