transport/tcp/AxisListener.java

[Sam Ruby <ru...@nc.rr.com>]

I've finally taken a look at the tcp transport contribution.

What does transport/tcp/AxisListener.java do that
transport/http/SimpleAxisServer.java doesn't do better?

Re: transport/tcp/AxisListener.java

[Rob Jellinghaus <ro...@unrealities.com>]

At 07:45 PM 5/27/2001 -0400, Sam Ruby wrote:
>If TCP was HTTP-MIME, I would be happy.  But you added back in content
>length.  And I presume that you would need something for sessions.  And then
>there is SOAPAction...

Yes, it's an ugly road of reinvention.  I thought I could get away without
any headers at all.  Silly me.

Originally I wanted to be able to interoperate with a rumored (by Glen :-)
TCP .NET binding.  No details emerged on that one though.  There is a TCP
binding for SOAP::Lite but interop with it is a low priority.

>If what you really wanted was a simple, standalone server that you could
>play with, then I totally agree.  Lets just converge on one.

Okey doke, and I'm down with it being SimpleAxisServer.  I was already
planning to create some functional tests with SimpleAxisServer as well (it
should be fairly easy now that the Ant groundwork is laid).

I may not take out the TCP code quite yet but nor will I do anything more
with it.

Cheers,
Rob

Re: transport/tcp/AxisListener.java

[Sam Ruby <ru...@nc.rr.com>]

Rob Jellinghaus wrote:
>
> Not much :-)  I didn't realize that SimpleAxisServer existed at the time I
> was doing the transport work.  You're right (you don't say this, but
you're
> thinking it) that we likely could do away with the TCP transport if we
> decided it wasn't worth supporting.
>
> It did enable me to factor out a lot of the transport-specific stuff in
> Call and Message, however, which will pay dividends when implementing SMTP
> (coming soon).  So it was still a worthwhile endeavor.

If TCP was HTTP-MIME, I would be happy.  But you added back in content
length.  And I presume that you would need something for sessions.  And then
there is SOAPAction...

So, if what we end up with is a is a competitor to MIME, then I'm not happy.
Why reinvent?

If what you really wanted was a simple, standalone server that you could
play with, then I totally agree.  Lets just converge on one.

Re: transport/tcp/AxisListener.java

[Rob Jellinghaus <ro...@unrealities.com>]

At 05:05 PM 5/27/2001 -0400, Sam Ruby wrote:
>I've finally taken a look at the tcp transport contribution.
>
>What does transport/tcp/AxisListener.java do that
>transport/http/SimpleAxisServer.java doesn't do better?

Not much :-)  I didn't realize that SimpleAxisServer existed at the time I
was doing the transport work.  You're right (you don't say this, but you're
thinking it) that we likely could do away with the TCP transport if we
decided it wasn't worth supporting.

It did enable me to factor out a lot of the transport-specific stuff in
Call and Message, however, which will pay dividends when implementing SMTP
(coming soon).  So it was still a worthwhile endeavor.

Cheers,
Rob