Indeed so. If some non-standardized protocol is out in the wild, and
acquiring new features in a haphazard manner, and if it contains at least
some features of merit, then standardizing it may well enable some control
of the situation (e.g. by deprecating the worst features, clarifying
intentions, resolving inconsistencies, discouraging competitive protocols,
etc.).
and sometimes we do this. HTTP, NNTP, SSL -> TLS all come to mind.
it sort of depends on how close the protocol or feature is in practice
to the behavior that is desirable. naturally if you can make the standard
sane while still making it more-or-less compatible with existing practice,
that's very useful in getting the standard deployed without significant
transition pain.
mainly what I disagree with is the idea that *because* something is in
widespread use then it should be standardized, and the notion that widespread
deployment of a protocol before standardization is an inherently good idea.
Keith