Re: http+srv worth its own URI scheme? (ISSUE-49 schemeProtocols-49)

On Feb 24, 2009, at 9:06 AM, Tim Berners-Lee wrote:

> Hmm. What is the overhead if every web browser does an SRV record  
> lookup?

Close to irrelevant compared to the HTTP transaction. It is very  
comparable to the use of CNAME records in DNS

> Does it do it at the same time as the A record lookup?

The DNS server, like CNAME, can look at the SRV record and return the  
associated A or AAAA records in the Answers section of the response  
the the SRV query. This is what many DNS servers do today for CNAME.  
If they don't do this, typically the client would need to a second DNS  
query to resolve the indirection, same as CNAME.

Normal DNS caching still applies so often all these responses are  
returned from a very close cache. When used in a example like facebook  
calling it, it's going to be the facebook server that does the DNS and  
and gets it it out of facebooks DNS cache.

> Does it wait to get back no SRV record before trying an A record  
> lookup?

Clients can do them in parallel if they want. I doubt people will  
bother to do that.

> Does this slow down HTTP operations at all?
> (What is the effect on the economy of slowing down an HTTP lookup by  
> each nanosecond?)
Uh, not much, otherwise people would get rid of DNS for redirections  
that did not need and just use IP addresses for stuff that needs to be  
fast or encode the http/html in a way that was smaller and faster to  

> If one *were* to introduce the hassle of a new URI scheme then I  
> would prefer to do something I had always wanted to do and remove  
> the difference between the . and the /
> as in:
> web:org/ietf/internet-drafts/draft-jennings-http-srv-02.txt
> where the algorithm would be to look for A record for  
> or an SRV for,
> or failing that the same up ( or SRV for org.) or down  
> (record for or an SRV for internet- 
> the tree.

That is very elegant but much more than I was trying to bite off in  
this small draft. One questions about this, if the client knew that it  
wanted the identity it was connecting to to be authenticated with a  
certificate with a subject alt name of web:org/ietf but was unwilling  
to accept web:org/ietf/internet-drafts, do you have a prosed syntax to  
do that?

> I think
> makes the mistake of introducing a very general and potentially  
> useful protocol but designing it for the case of new API  
> applications only is sub-optimal.
> I agree the overhead of a new URI scheme is too great.

Two questions, first when you talk about the overhead of introducing a  
new URI scheme, are you talking about the difficulty of getting this  
introduced and used on general web browsers for general web use? I  
agree that would be hard.

Second question. I'm just not sure how to take your above sentences.  
Are you saying this is good, but we need something far more general,  
but the far more general things will not happen because the overhead  
is too high, so we should not do the the more general thing? Does this  
imply we should do something simpler or do nothing at all?

>  The cost of https: as a separate scheme has I think been great too,  
> although I suspect one can argue a need for https: in way you can't  
> for http+srv.

I agree https has widespread usage and is more important than http+srv  
would be, but I don't understand what you mean by the cost of https as  
a separate scheme has been too great.  Can you educate me a bit on  
your view on https so I understand the parallel? Thanks, Cullen

> Tim
> On 2009-02 -23, at 13:32, Dan Connolly wrote:
>> srv records for web servers seems like a fine idea,
>> long overdue; but I'm skeptical that it's worth a new URI scheme.
>> I'm interested to know whether Noah and other TAG members
>> have other opinions.
>> cf. ISSUE-49 schemeProtocols-49
>> Begin forwarded message:
>>> From: Mark Nottingham <>
>>> Date: 23 February 2009 07:22:43 CEST
>>> To: HTTP Working Group <>
>>> Subject: Fwd: I-D Action:draft-jennings-http-srv-01.txt
>>> Archived-At:
>> <
>>> FYI.
>>> Begin forwarded message:
>>>> From:
>>>> Date: 23 February 2009 3:45:01 PM
>>>> To:
>>>> Subject: I-D Action:draft-jennings-http-srv-01.txt
>>>> Reply-To:
>>>> A New Internet-Draft is available from the on-line Internet-Drafts
>>>> directories.
>>>>     Title           : DNS SRV Records for HTTP
>>>>     Author(s)       : C. Jennings
>>>>     Filename        : draft-jennings-http-srv-01.txt
>>>>     Pages           : 7
>>>>     Date            : 2009-02-22
>>>> This document specifies a new URI scheme called http+srv which  
>>>> uses a
>>>> DNS SRV lookup to locate a HTTP server.  The http+srv scheme  
>>>> operates
>>>> in the same way as an http scheme but instead of the normal DNS A
>>>> record lookup that a http scheme would use, it uses an DNS SRV
>>>> lookup.  This memo also defines a https+srv scheme that operates in
>>>> the same was a an https URI but uses DNS SRV lookups.
>>>> The draft is being discussed on the list.
>>>> A URL for this Internet-Draft is:
>>>> Internet-Drafts are also available by anonymous FTP at:
>>>> Below is the data which will enable a MIME compliant mail reader
>>>> implementation to automatically retrieve the ASCII version of the
>>>> Internet-Draft.
>>> Content-Type: text/plain<BR>Content-ID:
>> &lt;
>>> &gt;<BR><BR>
>>>> _______________________________________________
>>>> I-D-Announce mailing list
>>>> Internet-Draft directories:
>>>> or
>> -- 
>> Dan Connolly, W3C
>> gpg D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Tuesday, 24 February 2009 18:40:13 UTC