W3C home > Mailing lists > Public > www-xkms@w3.org > September 2002

RE: XKMS tokens for use with SRV

From: Eastlake III Donald-LDE008 <Donald.Eastlake@motorola.com>
Date: Thu, 19 Sep 2002 15:20:55 -0400
Message-ID: <19CD0E423FC1D611893500508B6F0B9C0AC5F2@ma07exm01.dma.isg.mot.com>
To: "'www-xkms@w3.org'" <www-xkms@w3.org>

Strikes me that one part of my message wasn't clear due to the ellipsis. I meant three levels of service label in DNS. Like
	_XKMS-FOO-BAR._HTTP._TCP.example.com

Donald

-----Original Message-----
From: Eastlake III Donald-LDE008 
Sent: Thursday, September 19, 2002 11:59 AM
To: XKMS
Subject: RE: XKMS tokens for use with SRV

Joseph,

Well, it is a little unclear just how to handle a service built on HTTP with SRV but I'd be inclined to think it would be something like _XKMS..._HTTP._TCP.example.com.

I think the theory with SRV was that you didn't need to register anything with IANA because the "service name" would automatically be assigned when the port was assigned, as in the "key word" column of http://www.iana.org/assignments/port-numbers. So all you need to do is prefix an underscore. But that obviously doesn't work with higher level services not directly implemented on a port.

A reason for not just punting this is that "locate server via SRV" occurs very early in the use of the service so you really want to have it defined and implemented from the beginning or else you have backward compatibility problems and a lot of resistance to adding it. For example, this was not done with HTTP so you really can't add the requirement to use SRV to find HTTP servers at this late date. (At least HTTP has re-direct responses.)

If we want the facility, I suggest keeping SRV usage to the XKMS spec and I could whip up an ID on using SRV with higher level protocols/services.

Donald

PS: So if you have a // type URI that specifies a port and you use SRV to convert the host name in that URI to an address and port and the ports disagree, which dominates?

-----Original Message-----
From: Joseph Reagle [mailto:reagle@w3.org]
Sent: Wednesday, September 18, 2002 12:48 PM
To: XKMS
Subject: XKMS tokens for use with SRV

http://www.w3.org/2001/XKMS/Drafts/XKMS/xkms-issues-list#issue-13
>13 Clarification/ Phill Hallam- Baker, Joseph Reagle [ Sept 2002 F2F]
>[Part I - 2002/08/01]
>
>The SRV prefix for an XKMS service in DNS should be specified here once it 
>is established. Make changes as described.  [ Sept 2002F2F]

I have briefly investigated the registration of SRV service tokens for XKMS. 
SRV allows a single server (e.g., example.com) to delegate services to 
specific services. An example from from RFC2782:
   _Service._Proto.Name TTL Class SRV Priority Weight Port Target
I'm no expert in this, I've never heard of it prior to this issue being 
raised.

My conclusion is:
1. This is a relatively new standards track document and it's not clear that 
there's any immediate requirement for this from XKMS. Presently, SIP is the 
only registered service.
  http://www.iana.org/assignments/sip-table
2. There's no easy registration form for such values:
  http://www.iana.org/protocols/forms.htm
We would need to publish some form of informational RFC, I don't know what 
the requirements for such a document are.
2. I'm not sure if SRV's are intended to be used with services on top of 
HTTP. There is no _Proto corresponding to HTTP, and to be more specific 
(e.g,. _TCP or _UDP) would be inappropriate. One might take the cross 
product and have a service for every single protocol binding, such as 
XKMS+TCP, XKMS+UDP, XKMS+SMTP, etc., but this is not that compelling to me 
and I still can't distinguish between those on HTTP, and one actually 
running directly on TCP.

Consequently I recommend we take no action and wait for those Web services 
with actual requirements to take the lead if ever considered necessary.
Received on Thursday, 19 September 2002 15:20:57 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:31:39 UTC