Re: weighting transports

[Andrew Newton <[e-mail]>]

On Apr 1, 2010, at 14:59, "Peter Saint-Andre" <[e-mail]>  
wrote:

> Sometimes a client needs to determine which transport or technology an
> application server prefers for a given service type. For examples, an
> email service might be provided via both POP and IMAP or a website  
> might
> be available via both TCP and SCTP.
> 
> The approach taken in draft-daboo-srv-email is to check multiple SRV
> records (here IMAP would be preferred):
> 
>      _imap._tcp     SRV  0 1 143 imap.example.com.
>      _pop3._tcp     SRV 10 1 110 pop3.example.com.
> 
> The approach taken in draft-yourtchenko-tran-announce-dns is to  
> define a
> new DNS RR (here SCTP would be preferred):
> 
>      _http.www      XPORT    5 _sctp
>                     XPORT    9 _tcp
> 
> I think  it would be good to come to rough consensus (if possible)  
> on a
> preferred approach in the Applications Area.
> 
> Peter

I glanced at both drafts and did not see a discussion about why neither RFC 3958 nor RFC 4848 are sufficient for the purposes being described.  Perhaps that is a discussion I missed elsewhere.

I believe Philip Hallam-Baker put forward a very similar SRV proposal years ago but don't recall what happened to it.

Personally, I think it is time for the IETF to consider a new record type to simplify this type of DNS discovery if NAPTR and SRV records are considered too cumbersome for this purpose.

-andy