Re: [pcp] DNS Indirection

["Dan Wing" <[e-mail]>]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of
> [e-mail]
> Sent: Thursday, February 17, 2011 11:43 PM
> To: [e-mail]
> Subject: [pcp] DNS Indirection
> 
> Dear all,
> 
> We would like to have a feedback about an open question we have in
> http://tools.ietf.org/html/draft-bpw-pcp-dhcp-02 (Section 5.2):
> 
>       [Ed.  Is there a value to consider a level of indirection (e.g.,
>       SRV)? (1) to use an arbitrary port number for PCP Server instead
>       of the default port, (2) detect whether a security channel is in
>       use (using the transport protocol)]
> 
> The current text assumes there is no indirection.

There is value in SRV.  But there is additional complexity and additional
failure modes.  To my knowledge the only (relatively) widely deployed
application that uses SRV in real life is XMPP.  Just last year, Cisco 
bumped into a bug where our devices borked SRV responses that were 
longer than expected.  (1) would be neat to have, but I am not convinced
we necessarily need it.  (2) can be done without SRV; lots of protocols
without SRV already negotiate or force upgrading to a secure channel.

I vote for K.I.S.S.  That is, don't do SRV.

-d