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

[Randy Presuhn <[e-mail]>]

Hi -

On 5/7/2019 2:50 PM, Matthew Kerwin wrote:
> On Wed., 8 May 2019, 06:56 Barry Leiba, <[e-mail]
> <mailto:[e-mail]>> wrote:
...
>     No, of course not, not today, not with the history we have.  But if
>     they had *from the beginning*, we wouldn't be where we are.  And
>     that's the point.
>
>     Barry
>
>
> If the Internet (and particularly the web) hadn't "mostly just worked"
> despite us doing it all wrong, it might not have taken off at all.

The very early days of SNMP were plagued by subtle and not-so-subtle
issues resulting from overly-generous application of the Postel
principle.  Yes, it allowed a multiple implementations to be quickly
fielded and the whole thing appeared to work - for a while.  But it
also allowed oversights in the specification (e.g. how big can an
integer be?) to be glossed over, resulting in several interoperability
issues.  Fortunately, these problems were recognized fairly early in
the life of that protocol, and through a combination of industry-
accepted test suites as well as more precise specifications the
worst of the problems were resolved in the various code bases.

If the IETF (and the people writing the code) had not dealt with
those issues as quickly as they did, we'd still be battling those
problems today.

Randy