Re: [Uri-review] draft-merrick-jms-uri-01.txt

[Eric Johnson <[e-mail]>]

Hi Mark,

Mark Baker wrote:

Hi Eric,

On 2/27/08, Eric Johnson <[e-mail]> wrote:
  

 Specifically, in the introduction, change this:

 "The "jms" URI scheme is used to designate a javax.jms.Destination
 object and an associated javax.jms.ConnectionFactory object, and
 optionally provide additional information concerning the way that the
 Destination object is to be used. Probably the most common, ..."

 To this:

 "The "jms" URI scheme is used to designate a javax.jms.ConnectionFactory
 object and associated javax.jms.Destination objects, and optionally
 provides additional information concerning the way that the Destination
 objects are to be used.  The operations which can be performed on the
 ConnectionFactory and Destination resources are defined by the JMS
 specification and so a detailed description is outside the scope of this
 specification,
    

You don't get off the hook that easily 8-)  The registration requires
documentation of the operations.
  

???

Section 2.4 of RFC 4395 says that we *should* provide documentation of the operations, not that we must.  However, I think that is a distraction from the core concern, since we do document the operations by reference.

I wonder if we're talking about the same kinds of operations though:
would the JMS operations be ones that could be invoked remotely?
Because that's what the registration template is asking for, e.g. HTTP
GET for http:, FTP RETR for ftp:, etc... A quick look at the JMS
Javadocs for ConnectionFactory and Destination doesn't reveal any
application layer operations such as those AFAICT.
  

I believe you're right, we do disagree on the "kind" of operations.  In your response above, you indicate that you're expecting an "application" layer operation, but Section 2.4 doesn't actually use that qualification.  While section 2.4 proposes HTTP as a model for a URI that provides such operations, I don't read it as excluding other models.  For a clear counterpoint, the current treatment of the "mailto" URI in RFC 2368 does not require the use of "SMTP", nor should it.  I can think of uses where email gets sent with IMAP, a local sendmail agent, MAPI, or an HTTP web form (new in Firefox 3, I believe).  The "send" operation, then, is clearly decoupled from the application layer.  On the "operation" front, our JMS proposal seems to be in good company.

-Eric Johnson

_______________________________________________
Uri-review mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/uri-review