From: Ed Gerck <egerck(_at_)NMA(_dot_)COM>
Message-ID: <3C4E2A69(_dot_)6B2200DC(_at_)nma(_dot_)com>
Date: Tue, 22 Jan 2002 19:13:46 -0800
To: Randy Bush <randy(_at_)PSG(_dot_)COM>
CC: Ed Gerck <egerck(_at_)NMA(_dot_)COM>, ietf(_at_)ietf(_dot_)org
References:
<5(_dot_)1(_dot_)0(_dot_)14(_dot_)2(_dot_)20020122085259(_dot_)03a98008(_at_)127(_dot_)0(_dot_)0(_dot_)1>
<v04220813b8735cc508d1(_at_)[192(_dot_)168(_dot_)1(_dot_)14]>
<039201c1a389$eb0ed770$0301000a(_at_)FATORA>
<3C4E0AC7(_dot_)993D52B5(_at_)nma(_dot_)com>
<E16TDIp-0003WY-00(_at_)roam(_dot_)psg(_dot_)com>
Randy Bush wrote:
Perhaps we all agree that standardization without some sort of conformance
verification is ineffective.
can you read this mail? consider what it took to get it to you. how did
that happen?
Randy:
Once upon a time when the Internet was something we discussed at the
campus cafeteria, I might have found your questions to be useful. Today,
when much of our economy and future is riding on Internet technology which
is put together with what amounts to be bail wire and chewing gum, I find your
questions to be still useful. After all, if we had really thought hard about
those
questions then we would very probably be in a better position.
No, the success of the Internet is due to the flexibility of the
protocols on which it is based. The whole concept of "really thinking
hard" and coming up with an 'optimal' design or something was the
death knell of OSI. It is only because a zillion things were thrown
into the Internet, most of them something a grad student could do as a
project, and what worked and interoperated survived, that the Internet
is successful. The Internet is designed so it works to keep it
together with bailing wire and chewing gum.
Laissez-faire conformance has not worked. Period.
Conformance testing has never been a significant part of the
Internet. It was considered critical in OSI and other protocols
suites. Most people consider the Internet to be a success story and
OSI a failure.
IMO, we need to realize that
conformance verification is a fundamental part of the standardization process
and provide ways for it to happen. Publicizing non-conforming implementations
in an RFC would be a way to shine light onto them.
Conformance testing is mostly an insiginficant waste of
effort. Interoperability testing is what is important. The success of
the Internet is proof of this.
Conformance does not have to come with enforcement, though -- if that is your
fear. Just make it clear and public. As this very discussion is doing in
referenc
e to
Microsoft and its broken implementation of MIME.
True, conformance != enforcement. But it is also true that conformance
!= interoperability. And, quite rightly, what customers generally care
about is interoperability because that is what makes networks
valuable.
Cheers,
Ed Gerck
Donald