[AXIS Gen] Take a look at TRLSOAP

[Yuhichi Nakamura <NA...@jp.ibm.com>]

Axis contributers,
I am very concerning that most of you have not looked
at TRLSOAP yet.  I think that it is the only functional codebase
which implements our chaining idea.  Off course, class and interface
names are quite different from ones we are discussing, but
you could see how header handling chain works even with
intermediary.  In summary, similar concept is implemented in
a different way.

It is not a good idea to discuss Axis architecture without knowing
similar "functional" implementation.  Although I have been discussing
Axis, I have not seen big problem in TRLSOAP.  So, please, please
take a look at TRLSOAP.

After some of you guys look at it, I will raise two key questions:

1. Is a single Chainable API reasonable?
I still think that HeaderHanlder and others (ChainContainer, Transport
Listener) should be distinguished.  I am considering how we can
compare the two approaches.

2. Is managment stuff including Registry a core of engine?
As you can see from TRLSOAP, it does not management stuff, but you
"can" develop SOAP applications.  Our abstract API defines "minimum"
configuration API (called setNextConsumer), and we use it to
configure chains.  As a core, that is enough.  Management stuff
would be developed on top of the core, and invoke the API in more
smart way.  In this sense, it is an extention (but importnat) part of
engine.

Anyway, please, please, please take a look at TRLSOAP.
Thank you very much in advance.

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

RE: [AXIS Gen] Take a look at TRLSOAP

[James Snell <js...@lemoorenet.com>]

Yuhichi,

Actually, I've dug into it quite a bit and like what you've done.  I've
taken a stab at answering your questions below:

> -----Original Message-----
> From: Yuhichi Nakamura [mailto:NAKAMURY@jp.ibm.com]
> Sent: Saturday, January 06, 2001 3:45 PM
> To: soap-dev@xml.apache.org
> Subject: [AXIS Gen] Take a look at TRLSOAP
>
>
>
> Axis contributers,
> I am very concerning that most of you have not looked
> at TRLSOAP yet.  I think that it is the only functional codebase
> which implements our chaining idea.  Off course, class and interface
> names are quite different from ones we are discussing, but
> you could see how header handling chain works even with
> intermediary.  In summary, similar concept is implemented in
> a different way.
>
> It is not a good idea to discuss Axis architecture without knowing
> similar "functional" implementation.  Although I have been discussing
> Axis, I have not seen big problem in TRLSOAP.  So, please, please
> take a look at TRLSOAP.
>
> After some of you guys look at it, I will raise two key questions:
>
> 1. Is a single Chainable API reasonable?
> I still think that HeaderHanlder and others (ChainContainer, Transport
> Listener) should be distinguished.  I am considering how we can
> compare the two approaches.
>

I think that the single chainable API is very reasonable.  Some of my own
testing in this regard has shown quite a bit of success.  My goal is to
produce a shell implementation in C++ within about two weeks but we'll see.

> 2. Is managment stuff including Registry a core of engine?
> As you can see from TRLSOAP, it does not management stuff, but you
> "can" develop SOAP applications.  Our abstract API defines "minimum"
> configuration API (called setNextConsumer), and we use it to
> configure chains.  As a core, that is enough.  Management stuff
> would be developed on top of the core, and invoke the API in more
> smart way.  In this sense, it is an extention (but importnat) part of
> engine.
>

This is pretty much exactly what we're thinking Yuhichi, the management
stuff is not part of the core execution engine itself. Stuff like the WSMA
will be completely orthogonal to the core engine.

> Anyway, please, please, please take a look at TRLSOAP.
> Thank you very much in advance.
>
> Yuhichi Nakamura
> IBM Research, Tokyo Research Laboratory
> Tel: +81-46-215-4668
> FAX: +81-46-215-7413
>
>