You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@tuscany.apache.org by Luciano Resende <lu...@gmail.com> on 2010/02/26 19:50:32 UTC

Service Endpoint URI and it's behavior in Embedded versus Hosted Environment

Kelvin has identified a issue with store-webapp [1], and after some
investigation I  tried to summarize the current behavior in the
following wiki page [1]

Let's use this thread  to agree on what we think should be the proper
behavior, and then try to have a final solution to this issue, as
similar issues had come multiple times in the past [3][4]

[1] http://www.mail-archive.com/user@tuscany.apache.org/msg02268.html
[2] http://cwiki.apache.org/confluence/display/TUSCANYWIKI/Embedded+versus+Hosted+web+app+uri
[3] http://www.mail-archive.com/dev@tuscany.apache.org/msg10134.html
[4] http://www.mail-archive.com/dev@tuscany.apache.org/msg10284.html

-- 
Luciano Resende
http://people.apache.org/~lresende
http://twitter.com/lresende1975
http://lresende.blogspot.com/

Re: Service Endpoint URI and it's behavior in Embedded versus Hosted Environment

Posted by Luciano Resende <lu...@gmail.com>.

On Fri, Feb 26, 2010 at 2:13 PM, Raymond Feng <en...@gmail.com> wrote:
> Hi,
>
> Using a table is a good approach to document how the physical URI is
> constructed for a binding. I suggest that we add a few things:
>
> * A column for the base URI of the binding
> * Different rows to show the following:
>
> <binding.xyz/> <!-- no @uri -->
> <binding.xyz uri="http://example.com/ABC"/> <!-- absolute @uri -->
> <binding.xyz uri="ABC"/> <!-- relative URI not starting with / -->
> <binding.xyz uri="/ABC"/> <!-- relative URI starting with / -->
>

I have started a table in the expected behavior section [1], do we
thing we need any other info on that table ?

[1] http://cwiki.apache.org/confluence/display/TUSCANYWIKI/Embedded+versus+Hosted+web+app+uri

-- 
Luciano Resende
http://people.apache.org/~lresende
http://twitter.com/lresende1975
http://lresende.blogspot.com/

Re: Service Endpoint URI and it's behavior in Embedded versus Hosted Environment

Posted by Raymond Feng <en...@gmail.com>.

Hi,

Using a table is a good approach to document how the physical URI is 
constructed for a binding. I suggest that we add a few things:

* A column for the base URI of the binding
* Different rows to show the following:

<binding.xyz/> <!-- no @uri -->
<binding.xyz uri="http://example.com/ABC"/> <!-- absolute @uri -->
<binding.xyz uri="ABC"/> <!-- relative URI not starting with / -->
<binding.xyz uri="/ABC"/> <!-- relative URI starting with / -->

Thanks,
Raymond
--------------------------------------------------
From: "Luciano Resende" <lu...@gmail.com>
Sent: Friday, February 26, 2010 10:50 AM
To: "tuscany-dev" <de...@tuscany.apache.org>
Subject: Service Endpoint URI and it's behavior in Embedded versus Hosted 
Environment

> Kelvin has identified a issue with store-webapp [1], and after some
> investigation I  tried to summarize the current behavior in the
> following wiki page [1]
>
> Let's use this thread  to agree on what we think should be the proper
> behavior, and then try to have a final solution to this issue, as
> similar issues had come multiple times in the past [3][4]
>
> [1] http://www.mail-archive.com/user@tuscany.apache.org/msg02268.html
> [2] 
> http://cwiki.apache.org/confluence/display/TUSCANYWIKI/Embedded+versus+Hosted+web+app+uri
> [3] http://www.mail-archive.com/dev@tuscany.apache.org/msg10134.html
> [4] http://www.mail-archive.com/dev@tuscany.apache.org/msg10284.html
>
> -- 
> Luciano Resende
> http://people.apache.org/~lresende
> http://twitter.com/lresende1975
> http://lresende.blogspot.com/

Re: Service Endpoint URI and it's behavior in Embedded versus Hosted Environment

Posted by Luciano Resende <lu...@gmail.com>.

On Fri, Feb 26, 2010 at 11:35 AM, Simon Laws <si...@googlemail.com> wrote:
> So this seems to come down to how we interpret "/Catalog" in two
> different places.
>
> 1/ In javascript
> 2/ In an SCA (JSONRPC) binding URI
>
> I think it should be interpreted, in both of these places, in a way
> that is consistent with the specification of relative URIs in RFC2396
> [1].
>
> In the browser /Catalog will be interpreted as a root relative URL. In
> your case you have http://localhost:8080/Catalog.
>

Yes, this is the browser expectation.

> In the SCA binding we are indicating that we want the Catalog service
> made available at the URI "/Catalog". In lieu of a absolute path
> wouldn't we expect the runtime to interpret this a URI relative to the
> base URL with no path element, i.e. http://localhost:8080/Catalog?
>

Yes, that is my expectation as well, but it's not what's happening
when using host webapp in 2.x, where a service with URI /Catalog is
being registered relatively to the web application context root and is
available as http://localhost:9090/sample-store-webapp/Catalog

> Another way to look at this is to ask  how would you configure the
> binding if you did want the service to appear below the context root
> (which is possibly a reasonable, and safer, thing to do)?
>

This is a good question, and seems like a reasonable thing,
particularly if we want to scope the service to your application and
have multiple different versions of the same service provided form
different applications.... but if we can make the scenarios above
working, then, as long as the binding uri is properly configured,
things should work one way or another.

> If the binding uri were "Catalog" then the current code I believe adds
> the structural URI in front of the relative URI giving something like.
>  http://localhost:8080/context/component/service/Catalog. Which seems
> OK to me. In this case we've just constructed the base URI to which
> the binding URI is considered to be relative. As long as users
> understand this then can get any particular effect by manipulating the
> relative URI that they specify.
>

Yes, correct...
>
> [1] http://www.w3.org/TR/xmlschema-2/#RFC2396
>



-- 
Luciano Resende
http://people.apache.org/~lresende
http://twitter.com/lresende1975
http://lresende.blogspot.com/

Re: Service Endpoint URI and it's behavior in Embedded versus Hosted Environment

Posted by Simon Laws <si...@googlemail.com>.

So this seems to come down to how we interpret "/Catalog" in two
different places.

1/ In javascript
2/ In an SCA (JSONRPC) binding URI

I think it should be interpreted, in both of these places, in a way
that is consistent with the specification of relative URIs in RFC2396
[1].

In the browser /Catalog will be interpreted as a root relative URL. In
your case you have http://localhost:8080/Catalog.

In the SCA binding we are indicating that we want the Catalog service
made available at the URI "/Catalog". In lieu of a absolute path
wouldn't we expect the runtime to interpret this a URI relative to the
base URL with no path element, i.e. http://localhost:8080/Catalog?

Another way to look at this is to ask  how would you configure the
binding if you did want the service to appear below the context root
(which is possibly a reasonable, and safer, thing to do)?

If the binding uri were "Catalog" then the current code I believe adds
the structural URI in front of the relative URI giving something like.
 http://localhost:8080/context/component/service/Catalog. Which seems
OK to me. In this case we've just constructed the base URI to which
the binding URI is considered to be relative. As long as users
understand this then can get any particular effect by manipulating the
relative URI that they specify.

Regards

Simon

[1] http://www.w3.org/TR/xmlschema-2/#RFC2396