headers
Eric Johnson | 17 Sep 17:40
Favicon

Request for review: jms IRI - draft-merrick-jms-iri-00.txt

On behalf of a small collection of companies, I'm writing with the hope
that I'll collect feedback on the attached proposed IRI scheme for JMS
connectivity.

Any comments you can provide that would help adoption of this as a
full-fledged RFC are most welcome.  We working on and aware of
downstream specifications that will want to use some kind of URI scheme
for JMS communications, and we'd like to come up with as complete a
scheme specification as we can.

Thanks in advance.

-Eric Johnson



Network Working Group                                            D. Xiao
Internet-Draft                                                       BEA
Intended status: Informational                                R. Merrick
Expires: March 15, 2008                                              IBM
                                                              G. Daniels
                                                                Progress
                                                                T. Frank
                                                             Software AG
                                                              E. Johnson
                                                                   TIBCO
                                                      September 12, 2007

              IRI Scheme for Java(tm) Message Service 1.0
(Continue reading)

Permalink | Reply |
headers
Mark Baker | 17 Sep 19:45
Favicon
Gravatar

Re: Request for review: jms IRI - draft-merrick-jms-iri-00.txt

On 9/17/07, Eric Johnson <eric <at> tibco.com> wrote:
>    jms-iri = "jms:" jms-variant ":" jms-dest
>        [ '?' param  *([ '&' param ]) ]

That is incompatible with the generic syntax, which is;

  scheme ":" hier-part [ "?" query ]

I suspect "jms-jndi" is what you intended, but that's a URI scheme on
its own and so should be registered as such.  You can't just register
"jms-" as a prefix though.

But I'm really having a hard time understanding what this is trying to
accomplish.  IIRC, JMS is just an API, as is JNDI.  Knowing that a
service is implemented using either of these APIs tells me nothing
about how I might go about interacting with it over a network, which
is probably the most important practical role of a URI scheme.  If,
say, the service used AMQP behind the JMS API, then I could very well
see the value in an "amqp" URI scheme, because then the service
identifier would be exposing sufficient information for somebody to
interact with it.  But just JMS or JNDI?  That's insufficient
information.

Am I missing something perhaps?

Mark.

>
>    jms-variant = isegment-nz-nc
>
(Continue reading)

Permalink | Reply |
headers
Eric Johnson | 18 Sep 21:37
Favicon

Re: Request for review: jms IRI - draft-merrick-jms-iri-00.txt

Hi Mark,

Thanks for your feedback.  Some responses below.

Mark Baker wrote:
> On 9/17/07, Eric Johnson <eric <at> tibco.com> wrote:
>   
>>    jms-iri = "jms:" jms-variant ":" jms-dest
>>        [ '?' param  *([ '&' param ]) ]
>>     
>
> That is incompatible with the generic syntax, which is;
>
>   scheme ":" hier-part [ "?" query ]
>   
We have done our best to make sure that the syntax does, in fact,
conform to the syntax of an IRI.  We're expressing the syntax in a way
that is natural for identifying the concepts related to JMS, rather than
starting from the generic IRI syntax and adding a lot of text describing
the limitations on the uses of generic syntax.  Hopefully we got it right?

I allege (without formal proof at the moment, as I'm not quite familiar
with how one would go about proving this in ABNF syntax - I'll have to
dust off my compiler theory book, unless someone can point me at a
better reference), that the proposed JMS IRI syntax is a strict subset
of the IRI syntax.

Informally, I observe that the generic syntax you point at has three
parts, scheme, hier-part, and query.  I don't think we have any
disagreement about the proposed "scheme" and "query" parts being
(Continue reading)

Permalink | Reply |
headers
Mark Baker | 18 Sep 22:26
Favicon
Gravatar

Re: Request for review: jms IRI - draft-merrick-jms-iri-00.txt

On 9/18/07, Eric Johnson <eric <at> tibco.com> wrote:
> We have done our best to make sure that the syntax does, in fact,
> conform to the syntax of an IRI.  We're expressing the syntax in a way
> that is natural for identifying the concepts related to JMS, rather than
> starting from the generic IRI syntax and adding a lot of text describing
> the limitations on the uses of generic syntax.  Hopefully we got it right?

Oops, my bad.  I'm not sure what I was thinking when I looked at it.
The syntax is fine.

> > I suspect "jms-jndi" is what you intended, but that's a URI scheme on
> > its own and so should be registered as such.  You can't just register
> > "jms-" as a prefix though.
> >
> This is an interesting perspective that we admittedly haven't explored
> in detail.  I think the answer you come to in part comes from the
> analogies you make to other existing solutions.  We're thinking that the
> "jms" scheme identifies a JMS endpoint.  As such, it is far more useful
> to have a single "jms" scheme that allows multiple variants in how you
> discover that endpoint, rather than multiple schemes, which conceptually
> refer to the same thing.  Since our multiple variants can be used in an
> application in much the same way, it made sense to us to push the
> variant part further into the IRI.

It's been eons since I've used JMS, but what exactly is a "JMS
endpoint" nowadays?  It used to be an implementation detail.  i.e.
using my AMQP example, if party A wishes to send an AMQP message to
party B, it only cares whether party B speaks AMQP, not whether it
uses JMS.  Is that still the case?
(Continue reading)

Permalink | Reply |
headers
Eric Johnson | 19 Sep 22:37
Favicon

Re: Request for review: jms IRI - draft-merrick-jms-iri-00.txt

Hi Mark,

Again, thanks for feedback.  Turns out I had a few moments on unclarity
in my thinking and writing.

Certainly did not help the discussion that I confused AMQP with
something else, which I discovered with your response.  Worse yet, my
confusion inappropriately associated AMQP with offerings from IBM.  I'll
try below to reset the discussion around AMQP.

Now on to specific points.

Mark Baker wrote:
[snip]

> It's been eons since I've used JMS, but what exactly is a "JMS
> endpoint" nowadays?  It used to be an implementation detail.  i.e.
> using my AMQP example, if party A wishes to send an AMQP message to
> party B, it only cares whether party B speaks AMQP, not whether it
> uses JMS.  Is that still the case?

More precisely, the JMS term is a "Destination".  In JMS terms, a
"Destination" can be either a "Queue" or a "Topic".

Now that I have more of a clue, I think AMQP is an orthogonal question
here.  JMS, as an API, can be implemented with the help of HTTP, BEEP,
or custom packets over "raw" TCP/IP.  Likewise, JMS could be implemented
over AMQP.

To turn your example around, if party A wishes to send a JMS message to
(Continue reading)

Permalink | Reply |
headers
Martin Duerst | 18 Sep 06:29

Re: Request for review: jms IRI -draft-merrick-jms-iri-00.txt

Would it be possible to actually submit this 'draft', rather
than just send something that looks like an Internet-Draft?

Thanks,    Martin.

At 00:43 07/09/18, Eric Johnson wrote:
>On behalf of a small collection of companies, I'm writing with the hope
>that I'll collect feedback on the attached proposed IRI scheme for JMS
>connectivity.
>
>Any comments you can provide that would help adoption of this as a
>full-fledged RFC are most welcome.  We working on and aware of
>downstream specifications that will want to use some kind of URI scheme
>for JMS communications, and we'd like to come up with as complete a
>scheme specification as we can.
>
>Thanks in advance.
>
>-Eric Johnson
>
>
>
> > >Network Working Group                                            D. 
>Xiao >Internet-Draft                                                       
>BEA >Intended status: Informational                                R. 
>Merrick >Expires: March 15, 2008                                             
> IBM >                                                              G. 
>Daniels >                                                                
>Progress >                                                                T. 
>Frank >                                                             Software
(Continue reading)

Permalink | Reply |
headers
Eric Johnson | 18 Sep 18:32
Favicon

Re: Request for review: jms IRI -draft-merrick-jms-iri-00.txt

Hi Martin,

We'd like an informal review first, which is why we submitted it the way
we have at the moment.  Perhaps I should have mentioned that we were
explicitly looking at an informal review at this point, rather than a
formal one?

Once we get some feedback, hopefully we can take the concerns raised
back to our working group and iron out the issues before we do a formal
submission.  Does that work?

-Eric Johnson

Martin Duerst wrote:
> Would it be possible to actually submit this 'draft', rather
> than just send something that looks like an Internet-Draft?
>
> Thanks,    Martin.
>
> At 00:43 07/09/18, Eric Johnson wrote:
>   
>> On behalf of a small collection of companies, I'm writing with the hope
>> that I'll collect feedback on the attached proposed IRI scheme for JMS
>> connectivity.
>>
>> Any comments you can provide that would help adoption of this as a
>> full-fledged RFC are most welcome.  We working on and aware of
>> downstream specifications that will want to use some kind of URI scheme
>> for JMS communications, and we'd like to come up with as complete a
>> scheme specification as we can.
(Continue reading)

Permalink | Reply |
headers
Martin Duerst | 19 Sep 04:05

Re: Request for review: jms IRI -draft-merrick-jms-iri-00.txt

At 01:32 07/09/19, Eric Johnson wrote:
>Hi Martin,
>
>We'd like an informal review first, which is why we submitted it the way
>we have at the moment.  Perhaps I should have mentioned that we were
>explicitly looking at an informal review at this point, rather than a
>formal one?

That's not how the IETF is working. Submitting an Internet-Draft
isn't equivalent to a formal submission. It's perfectly okay
(and the usual way to do) to submit an Internet-Draft and ask
for informal comments. Indeed I don't know any case where the
first version of an Internet-Draft was a 'formal submission'
(but given the number of Internet-Drafts out there, I'm sure
there must be one or two such cases).

The process of publishing an Internet-Draft is separate from
any 'formal submission' steps. It's easily possible that you
submit a document as an Internet-Draft, and a few weeks later
decide to 'formally submit' it because you haven't received
any further comments.

Publishing an Internet-Draft is extremely easy. You already
have done all the formatting as far as I can see (except that
the line endings didn't get through to me in email). A mail
to the respective address is all you need.

Regards,    Martin.

>Once we get some feedback, hopefully we can take the concerns raised
(Continue reading)

Permalink | Reply |

Gmane