Re: New SASL capability for the Python client

2009-11-11 Thread Andrew Stitcher
On Wed, 2009-11-11 at 15:50 -0500, Ted Ross wrote:
 Full SASL authentication/encryption capability for the Python client was 
 added to the trunk at revision 834975.
 
 A new Python module qpidsasl implemented in C++ and wrapped for Python 
 using Swig was introduced.  This wrapper provides a generalized binding 
 to the Cyrus SASL library.  The Python client tries to import this 
 module.  If it cannot find it, it will revert to built-in capability 
 that only provides ANONYMOUS and PLAIN authentication mechanisms.

This would appear to not really be connected with qpid itself, but
rather be a (very) useful addition to the python (and ruby) libraries.

I'd say it would actually be better and more generally useful (for other
applications) to put this code in an entirely separate repository from
qpid and for it to distributed entirely separately from qpid. And so to
remove the qpid element of its name.

For Fedora (and the like you'd package it in packages called
python-sasl, ruby-sasl unless those names are already taken)

Have I missed something here that makes this actually specific to qpid?

Andrew



-
Apache Qpid - AMQP Messaging Implementation
Project:  http://qpid.apache.org
Use/Interact: mailto:dev-subscr...@qpid.apache.org



Re: New SASL capability for the Python client

2009-11-11 Thread Andrew Stitcher
On Wed, 2009-11-11 at 16:29 -0500, Andrew Stitcher wrote:
 On Wed, 2009-11-11 at 15:50 -0500, Ted Ross wrote:
  Full SASL authentication/encryption capability for the Python client was 
  added to the trunk at revision 834975.
  
  A new Python module qpidsasl implemented in C++ and wrapped for Python 
  using Swig was introduced.  This wrapper provides a generalized binding 
  to the Cyrus SASL library.  The Python client tries to import this 
  module.  If it cannot find it, it will revert to built-in capability 
  that only provides ANONYMOUS and PLAIN authentication mechanisms.
 
 This would appear to not really be connected with qpid itself, but
 rather be a (very) useful addition to the python (and ruby) libraries.
 
 I'd say it would actually be better and more generally useful (for other
 applications) to put this code in an entirely separate repository from
 qpid and for it to distributed entirely separately from qpid. And so to
 remove the qpid element of its name.

Specifically I'd add that its location in the source tree is not correct
in my mind - it is not really any part of the c++ implementation of the
amqp protocol and it is not a binding to a any qpid library so putting
it in cpp/bindings is more confusing than not.

It also adds to an already complex c++ build.

I'd prefer to see it moved to its own top level directory until we can
put it outside qpid altogether. Say in sasl.

Andrew



-
Apache Qpid - AMQP Messaging Implementation
Project:  http://qpid.apache.org
Use/Interact: mailto:dev-subscr...@qpid.apache.org



Re: New SASL capability for the Python client

2009-11-11 Thread Carl Trieloff

Andrew Stitcher wrote:

On Wed, 2009-11-11 at 16:29 -0500, Andrew Stitcher wrote:
  

On Wed, 2009-11-11 at 15:50 -0500, Ted Ross wrote:

Full SASL authentication/encryption capability for the Python client was 
added to the trunk at revision 834975.


A new Python module qpidsasl implemented in C++ and wrapped for Python 
using Swig was introduced.  This wrapper provides a generalized binding 
to the Cyrus SASL library.  The Python client tries to import this 
module.  If it cannot find it, it will revert to built-in capability 
that only provides ANONYMOUS and PLAIN authentication mechanisms.
  

This would appear to not really be connected with qpid itself, but
rather be a (very) useful addition to the python (and ruby) libraries.

I'd say it would actually be better and more generally useful (for other
applications) to put this code in an entirely separate repository from
qpid and for it to distributed entirely separately from qpid. And so to
remove the qpid element of its name.



Specifically I'd add that its location in the source tree is not correct
in my mind - it is not really any part of the c++ implementation of the
amqp protocol and it is not a binding to a any qpid library so putting
it in cpp/bindings is more confusing than not.

It also adds to an already complex c++ build.

I'd prefer to see it moved to its own top level directory until we can
put it outside qpid altogether. Say in sasl.
maybe create a util/sasl directory... I would not go to the effort of 
another package unless there is

a LOT of interest to do so

Carl.




Re: New SASL capability for the Python client

2009-11-11 Thread Ted Ross

On 11/11/2009 04:29 PM, Andrew Stitcher wrote:

On Wed, 2009-11-11 at 15:50 -0500, Ted Ross wrote:
   

Full SASL authentication/encryption capability for the Python client was
added to the trunk at revision 834975.

A new Python module qpidsasl implemented in C++ and wrapped for Python
using Swig was introduced.  This wrapper provides a generalized binding
to the Cyrus SASL library.  The Python client tries to import this
module.  If it cannot find it, it will revert to built-in capability
that only provides ANONYMOUS and PLAIN authentication mechanisms.
 

This would appear to not really be connected with qpid itself, but
rather be a (very) useful addition to the python (and ruby) libraries.

I'd say it would actually be better and more generally useful (for other
applications) to put this code in an entirely separate repository from
qpid and for it to distributed entirely separately from qpid. And so to
remove the qpid element of its name.

For Fedora (and the like you'd package it in packages called
python-sasl, ruby-sasl unless those names are already taken)

Have I missed something here that makes this actually specific to qpid?

Andrew

   
The library that holds the wrapper code is libsaslwrapper.so.  The 
Python binding I called qpidsasl but you are correct in saying that 
there is nothing qpid or messaging related in either the library or the 
module.  All the library does is provide an alternate API that doesn't 
rely on callbacks and is easily wrapped in scripting languages.  I'd be 
happy to move the whole thing to a top level directory under qpid if 
there's a consensus that that is the right thing to do.


I assume that we, as the Qpid community, don't have access to 
directories above qpid.  I just wanted to implement SASL for Python and 
Ruby, not start a whole new project.


-Ted


-
Apache Qpid - AMQP Messaging Implementation
Project:  http://qpid.apache.org
Use/Interact: mailto:dev-subscr...@qpid.apache.org