Re: [tae] New draft: announcing the supported transports via DNS

[Joe Touch <[e-mail]>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1



Dan Wing wrote:
>> Dan Wing wrote:
>>>> As you note, addresses are sometimes used for non-human 
>>>> purposes, and
>>>> with IPv6 they could be created on the fly - I wouldn't want 
>>>> to have to
>>>> wait to register them in the DNS vs exchanging them in-band.
>>> So, when you exchange them (in whatever 
>>> application-specific protocol 
>>> you're using to exchange those addresses, e.g., SDP for RTSP, SAP, 
>>> and SIP) using that application-specific protocol is the best place 
>>> to indicate which transport protocol is supported on the host. 
>> I might want to say "here, use this address, but use whatever 
>> transport the other end wants to use for reliable object 
>> transfer", which wouldn't have the information available to 
>> the protocol that indicates the address.
> 
> I don't understand; if you have an existing protocol to send
> an IPv4 or IPv6 address around, it can be extended to include
> a port (as most application developers have found necessary,
> as we can't have more than one HTTP server listening on port
> 80 behind an in-home NAT), and it can likewise be extended to
> say "TCP=yes, SCTP=yes, DCCP=no".

It can. Or it can be "here's an address and a port, you figure out what
transports are provided".

> You're saying such an extension is not possible / reasonable?

I'm saying it should not be strictly required.

> Do you have a proposal you could share on the mailing list or
> in an I-D?

No proposal; just trying to see if proposed solutions can handle cases I
consider useful.

Joe
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)

iEYEARECAAYFAkqz2E4ACgkQE5f5cImnZrs0LQCfePjjO8Fa7GYn627BH3pxn31b
FOEAn2/DF1X8mkmWmrYrOKd3sQVAwtGl
=0YqU
-----END PGP SIGNATURE-----