You are viewing a plain text version of this content. The canonical link for it is here.
Posted to soap-dev@xml.apache.org by Steve Graham <sg...@us.ibm.com> on 2001/01/01 15:08:10 UTC

RE: [Axis Gen] Comments on Design Doc.

Yuhichi:
I agree the approach we take at IBM is sometimes too slow.

In the Axis case, however,  we need to at least agree on the major
sub-systems (for lack of better word), agree on how the sub-systems
collaborate, agree on the APIs that mark the boundaries between the
components of the various subsystems and agree on who builds which
subsystem.  I would prefer the approach of building in parallel but all of
us working on complementary pieces.

Before we can get this level of parallelism, we need to agree on the
architecture and API.  The arch. overview is trying to lay the framework
for the architectural agreement and should also identify the major
subsystems.  A set of Java interface files should be sufficient to agree on
API (this should not take long to produce after the architecture is agreed
upon).  We then need to agree on who builds what, and get coding.

If there are suggestions for alternative approaches from anyone in the
community, please share them.

sgg

++++++++
Steve Graham
sggraham@us.ibm.com
(919)254-0615 (T/L 444)
<<Pithecanthropus Erectus>>
Emerging Internet Technologies
++++++++


"Yuhichi Nakamura" <NA...@jp.ibm.com> on 12/30/2000 12:49:02 AM

Please respond to soap-dev@xml.apache.org

To:   soap-dev@xml.apache.org
cc:
Subject:  RE: [Axis Gen] Comments on Design Doc.




Steve,
I am happy to hear that the purpose of the document to make sure
we agreed on the concepts in Section 3.
As for development, I am not sure if we should make a detailed desgin
first, then develop each module in parallell.  In IBM, we follow such
process
actually, but it always takes so much time to complet implementation.
My suggestion is that as soon as we agree on the basic concepts, then
each person or team develops a prototype.  If we have more than one
prototypes, we vote on them.
As you mentioned before, we would get the initial version in April if
we tried to agree on implementation details.  It is too late, I think.
Best regards,

Yuhichi Nakamura
IBM Research, Tokyo Research Laboratory
Tel: +81-46-215-4668
FAX: +81-46-215-7413


From: "Steve Graham" <sg...@us.ibm.com> on 2000/12/30 01:43

Please respond to soap-dev@xml.apache.org

To:   soap-dev@xml.apache.org
cc:
Subject:  RE: [Axis Gen] Comments on Design Doc.



The major purpose of the architecture overview was to make sure we agreed
on these concepts (section 3), and how they relate to each other (the "use
cases" outlined in sections 4 thru 9.

Once we have agreement on these concepts, an API agreement (design) can be
made, we can then parcel out the major modules of the architecture and
build in parallel.

sgg

++++++++
Steve Graham
sggraham@us.ibm.com
(919)254-0615 (T/L 444)
<<Pithecanthropus Erectus>>
Emerging Internet Technologies
++++++++


"Yuhichi Nakamura" <NA...@jp.ibm.com> on 12/28/2000 02:13:37 AM

Please respond to soap-dev@xml.apache.org

To:   soap-dev@xml.apache.org
cc:
Subject:  RE: [Axis Gen] Comments on Design Doc.




James,
I feel that we should agree on basic notions that are described with a
set of terminologies as in Section 3.2 and 3.3.  When I looked at
these sections, we have 4 or 5 enties for object structure,
and around 10 entities for processing.  Currently we have not
agreed on even this small set of terminologies.

Maybe, our eventual goal is to agree on this abstraction level.
Does it make sense?

Yuhichi Nakamura
IBM Research, Tokyo Research Laboratory
Tel: +81-46-215-4668
FAX: +81-46-215-7413


From: "James Snell" <js...@lemoorenet.com> on 2000/12/28 15:13

Please respond to soap-dev@xml.apache.org

To:   <so...@xml.apache.org>
cc:
Subject:  RE: [Axis Gen] Comments on Design Doc.



Yuhichi,

I think I can speak for Steve as well as myself in saying that we both
agree
that the architecture overview is bit more complex than it really needs to
be, and we're working on simplifying it now.  The finished product will
include ideas from both the doc Steve produced (which was based on notes he
took at the face to face in Washington and the document I produced which
was
basically intended at offering amendments to Steve's rendition.  We plan on
pinging the group for discussion and consensus on architecture once all of
the key members have returned and then moving forward on implementation.

- James

-----Original Message-----
From: Yuhichi Nakamura [mailto:NAKAMURY@jp.ibm.com]
Sent: Wednesday, December 27, 2000 10:08 PM
To: soap-dev@xml.apache.org
Subject: [Axis Gen] Comments on Design Doc.



I am still not sure if we should produce such detailed architecture design.
However, it is worthwhile to have some agreements on basic notions
described
in Section 3.  Here are my comments.
p.s. James, I just took a look at yours.  I will comment on it later.

Overall
- I would like to use terminology provided by XP as much as possible.
- I would like you to distinguish what are agreed, and what are just
  your thoughts.

3.2 Object Structure ...
-Message is OK.  It would be called Envelope instead.
-We may need a global entity like AxisContext, but I have no idea on
usecase.
-TranposrtContext seems too specific.  We may use MessageContext as
  an abstaction.  In addition, we may need SessionContext.  MessageContext
  has an information related to a "single" message path.  I assume that
  request/response is comprised of two message paths.  In that case,
  we may need SessionContext to share information between request
  and response message path.

3.3 Component Overview
- Are you claiming that all entities (3.3.2-3.3.13) are Chainable?   Which
are
  /are not Chainable?
- Why do you need Chain as tangible object?  It can be embedded in
  ChainContainer as a Chain object or just a Vector of Handlers.
  Chain can be encapsulated within ChainContainer.
- In terms of SOAP spec, Header Hanlder and Body Handler might be
   distinguished because different fault exceptions are thrown by them.
- As we are discussing in other mails, Router is not a part of
   Registry, I think.  (See our TRLSOAP)
- Dispatcher is ambiguous.  It should be RPCHandler or something.
- TransportSender/Listener seems to specialized for one way messaging.
   How do they work for request/response?
- Should we specify Serializer/Deserializer/Cannonicalizer here?  To me,
   they are components used in TransportSender/Listener.
- Why do you need ChainableRegistry here?  Isn't is a portion of
   ServiceRegistry?
In summary, I prefer to use the following terms:
    ChainContainer, HeaderHandler, BodyHandler, Router, RPCHandler,
    TransportListener, TransportSender, ServiceRegistry

More comments will be posted later.

Yuhichi Nakamura
IBM Research, Tokyo Research Laboratory
Tel: +81-46-215-4668
FAX: +81-46-215-7413


From: Yuhichi Nakamura/Japan/IBM@IBMJP on 2000/12/27 13:32

Please respond to soap-dev@xml.apache.org

To:   soap-dev@xml.apache.org
cc:
Subject:  [Axis Gen] How to Proceed?




Axis folks,
I just read through the design document placed at the following URL:
http://www.soap-wrc.com/axis/axis_arch_overview_v03.zip

Then, I came up with a question like:
    Should we produce a complete design document before the actual
    developement?
At F2F,  we agreed that we should have design discussion first,
reach an agreement on an architecture design, then start implementation
assigning some modules to someone.  I am not sure that this is the best
way to proceed Axis project.  To me, even design discussion seems to
last forever!

TRLSOAP implements some of key features discussed here, and we
want to improve it soon.  Should we wait for the complete design?
Maybe I am concerned that nothing is implemented, but
bunchs of documents are/will be produced.  Is there someone
who is implementing the design?

There are at least two ways to proceed this project:
1. Complete a design doc., then implement it (as we are doing now)
2. Start from a functional prototype (such as Idooxoap and TRLSOAP)
*Apache SOAP V1/2 took the second approach.

Anyway, I am worring about the slow progress of this project.
Regards,

Yuhichi Nakamura
IBM Research, Tokyo Research Laboratory
Tel: +81-46-215-4668
FAX: +81-46-215-7413