Re: [DNA-BOF] Modified Charter available.

[Thomas Narten <narten@us.ibm.com>]

> I keep on hearing about API being worrisome though.

The IETF is wary of APIs. For one thing, it is unclear that we really
have the expertise to do them correctly. There are other bodies that
traditionally have done APIs. E.g, even the IPv6 sockets API is only
informational, with the "real" standardization done by the Austin
Group. They took the RFC, but their version is the one that is
considered the official standard.

Also, APIs can easily be a tad OS specific. That makes it harder to
come up with one that is considered to be appropriate by all.

Having said that, the IETF have standardized some APIs, in cases where
we feel it is really important, but only in a few cases. GSSAPI comes
to mind.

> > |1) API definition (even the sockets API raises concern)
> > 
> > That is not entirely true. RFC 2960 (Standards Track) includes
> > nearly 15 pages "[sketching] the API which should exist at the
> > boundary between the SCTP and the SCTP user layers."

There is a big difference between "sketching the API" and defining a
standard. The former is quite useful and we've done it in a lot of
documents. But the sketch ends up being input to folks doing the
actual API, rather than being the API itself.

Thomas