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

[Carsten Bormann <[e-mail]>]

Hi Martin,

A hearty +1 to most of what you wrote.  Many IETFers now understand that it is important to prevent protocols from degenerating into soup, and we may be unique as an organization by actually started making that a part of our culture.  We should continue on this trajectory.

I think there is one detail missing from this dilemma:

>> — trying to specify every possible detail […]
>> — specifying the normal cases and general philosophy […]

In some parts of the specification, we actually can achieve the former without loss of the latter: Where we have formal description techniques (FDT) that are actually *usable*.

Admittedly, we are as far away from covering whole protocols with usable FDTs as we were 30 years ago when the misapplication of FDTs helped OSI fail.  But for those narrow parts where we have made progress, we should be embracing those usable FDTs, and should actively work on capability building.  We can already achieve good coverage of some areas where many implementation bugs previously tended to lurk.  And FDTs can be an effective bulwark against soupification.

Extending the (usable) reach of FDTs is research, and the new IRTF chair, Colin Perkins, has mentioned that FDTs are high on his agenda.  For IETF use, usability is key; as you say, a thousand pages of assembly language level detail without a big picture visible doesn’t cut it, and I’d add that is true independent of whether the “assembly language” is presented in stylized English or in an unusable FDT.

Grüße, Carsten