axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rob Jellinghaus <r...@unrealities.com>
Subject Re: TCP transport & functional test, w/o refactoring
Date Sat, 19 May 2001 02:31:53 GMT
It's not so much hesitation to refactor; I just wanted to send out a
submittable (even if suboptimal) snapshot ASAP.  Getting the functional
test working was higher priority than cleaning things up.

No worries, I'll be doing the refactoring next (mostly tomorrow), and will
send another zip when done (Monday at the latest).  Your point about not
checking in architecture-breaking code is well taken... this "only add
files" experiment is the last such experiment I'll do.

Cheers,
Rob

At 10:16 PM 5/18/2001 -0400, Doug Davis wrote:
>Rob,
>  I don't understand your hesitation about modifying the axis code.
>Having each transport create its own call/message/admin classes
>is wrong so I don't see why we would integrate the code you sent.
>While you're right having this code won't break anything, it does
>break the architecture - don't be shy about it - just modify whatever
>files you need in order to make things work they way they should.
>-Dug
>
>
>Rob Jellinghaus <robj@unrealities.com> on 05/18/2001 09:55:00 PM
>
>Please respond to axis-dev@xml.apache.org
>
>To:   axis-dev@xml.apache.org
>cc:
>Subject:  TCP transport & functional test, w/o refactoring
>
>
>
>I just got the Ant functional test case working with the new TCP transport.
> You can (or I can, anyway) do
>
>xml-axis/java> java org.apache.tools.ant.Main functional-tests
>
>and it will:
>
>- build Axis
>- build an Ant task (required for launching a server from within Ant)
>- launch an AxisListener (an Axis TCP server)
>- run a JUnit test case which sends a message to the server and validates
>the response
>- close the AxisListener
>
>Caveats:
>
>- The current test case should be something simple like echo, but it's not
>clear to me how EchoHandler works if there is no SOAPAction HTTP header
>(the TCP transport has no HTTP headers at all).  So the current
>AxisListener test case support is grottier than it should be.
>
>- The AxisListener also defines a hacky wire protocol to support
>client-sent ping and quit commands.
>
>- I have deliberately not yet touched ANY existing Axis files (except for
>build.xml).  I know that some of the classes in this zip (mainly TCPCall
>and TCPMessage) should not exist, or at least should be refactored w.r.t.
>HTTPCall and HTTPMessage.  I am going to do that next, but I wanted to send
>this off ASAP so Sam won't accuse me of being unable to check in for more
>than three days :-)
>
>In other words, this is submittable now without breaking anything (since no
>existing files are touched), but I will keep working on it whether you
>submit it now or not.  It's your call.
>
>Once I get the basic refactorings ({HTTP,TCP}{Call,Message,Dispatcher})
>done, I would like to start work on a proper architecture for configurable
>transports & listeners (a la Glen's ideas).  The current TCP transport (and
>transports yet unwritten) should be far more configurable and flexible, and
>I'd like to make it so.
>
>Let me know what you all think.
>Cheers,
>Rob
>
>
>
>
>
>
>
>
>


Mime
View raw message