Re: [arch-d] deprecating Postel's principle - considered harmful

[Keith Moore <[e-mail]>]

I'm another person who really doesn't want to see an RFC of the form 
"Jon Postel was wrong".   Part of the problem is that he was right in so 
many ways, that to single out in a particular way that he was "wrong" 
[*], conveys entirely the wrong impression.   Even if he was "wrong" [*] 
about the sentence most often attributed to him today.

[*] or lacking perfect foresight, or is being misunderstood, or taken 
out of context, or ...

Depending on where you mark the start of the Internet, it and the first 
protocols have existed for around forty years.   It's been around for so 
long, longer than most of its users have been alive, that it's easy to 
take for granted the accumulated wisdom that went into producing it - 
both Jon's contributions and others'. That's not to say that they got 
everything right, and today we're painfully aware of some of the 
shortcomings.    But that they got it working well enough to endure for 
40 years and counting, and to be able to evolve (to some extent) to 
continue to serve the whole world today, was a tremendous 
accomplishment.   Of course, different bits of wisdom become more (or 
less) applicable (or need to be interpreted in different contexts) at 
different points of evolution and scale.

I seriously doubt that Jon would have wanted us to be stuck in the past 
and stop learning.   But neither should we discard hard-won lessons of 
the past.

A lot of additional hard-won wisdom has been cited in this thread, and I 
think it could do a service to try to capture it and place it in the 
context of the earlier accumulated wisdom.

Keith

p.s.  I've often said that "the web" was optimized for deployability.   
Lots of "the web" was poorly designed, IMO, but enough of it was 
designed just well enough to make it attractive to users, and that very 
attractiveness is what has fueled the effort required to improve it.   A 
similar statement could be made of the Internet itself.  As an engineer, 
this bugs me a bit, but I think it's reality.   It's often said that the 
perfect is the enemy of the good, but maybe sometimes, the good is the 
enemy of the deployable.   Being liberal in what you accept is similar, 
in a sense, to optimizing for deployability.