> > but you have no way of knowing whether the app has made that
> > decision.
> > neither does the host.  we have a clear consensus to not change the
> > existing API behavior, which assumes stable addresses by
> > default, and this implies that an app can 'choose' a stable
> > address merely by not taking any special action.  this is as
> > it should be.
> 
> The IETF does not standardize APIs 

As a categorical statement, this is wrong.  The IETF can and does sometimes 
standardize APIs, it's just that the IETF emphasizes interoperability 
at the "bits on the wire" level rather than at the API level. this is both 
because experience indicates that API-level standards aren't sufficient to 
allow different implementations to interoperate, and because there's no need 
to insist that everyone use the same API - it actually degrades interoperability
to do so.

> and it's perfectly OK for different
> implementations to have different APIs with different assumptions about
> address stability. 

I agree with this part.  If you want to design a new API in which the
default behavior is to assign a temporary address and apps have to ask
for a stable address, I have no problem with that.  However, I have a big 
problem with this document saying it's okay to change the behavior of 
existing APIs to assign temporary addresses to apps that were designed
and coded to expect stable addresses.

> > host implementors have no business making such decisions.  or
> > at least, IETF shouldn't endorse such brain-damage if they do.
> 
> The host implementor *must* make a decision about what to do when there
> is no input to override the implementation's default behavior.

That's a cumbersome statement; it's like saying tha the host implementor
must have a default behavior when the default behavior isn't specified.

A simpler but equivalent statement is that there must be a default 
behavior for any option for which the API doesn't force an explicit choice.    
That's true on its face, but it really doesn't illuminate the quesiton.  

The real question is whether this document should endorse the idea that the 
host can change existing APIs in such a way that it may alter the behavior
of programs written to those APIs.  The only reasonable answer is "no".

Keith
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to