On Thu, 2012-07-12 at 23:26 +0200, Rob Godfrey wrote:
> There are two reasons why I think amqp-libraries@ is strictly superior
> to proton@.
> 
> Firstly, while you may consider "proton" as a distinct stream of work,
> I'm not sure how that works with something like the messaging API
> which fits in with your "quadrant" of APIs, but is clearly not part of
> Proton as we know it today.

Fitting into the matrix doesn't say much one way or the other. The
matrix is categorizing APIs according to user requirements. Lots of APIs
will fit in the matrix, even APIs developed elsewhere which clearly
wouldn't belong on any qpid mailing list.

The reason that the matrix is relevant to proton is that as the goal of
proton is to provide for the easiest possible integration of AMQP 1.0
into the widest possible range of applications, it needs to cater to all
four quadrants of that matrix, i.e. make it super easy to integrate with
an application or API that uses any one of those models.

>From that perspective, I see the future of the messaging API happening
in two places. Part of it I think will happen in the proton work stream
as I would like to see the future messaging API built on top of proton
and I think it should take as much advantage as possible of the
architecture proton has for delivering functionality into lots of
different languages, e.g. the swig bindings, the python/jython testing
infrastructure, etc. This part of it I think really constitutes filling
out proton's functionality in the upper right quadrant of the matrix,
and I see developers of the future messaging API participating in the
proton work stream discussions in the role of proton users.

The second part which I see happening outside the proton work stream is
really to do with the messaging APIs origins. It's core requirement from
the start has been to provide a way for users of the qpid brokers to
transition to 1.0 from older protocols. As such there are very likely
things that the implementation of the future messaging API will want to
do that may well be specific to the qpid brokers, and doing this is
perfectly in keeping with that original goal of the messaging APIs, but
is not consistent with the goal of proton which requires being broker
agnostic. These sorts of things I would expect to be part of the
discussion between the future messaging API developers and the future
messaging API users, which would really be a conversation that happens
on the current qpid lists.

> Secondly, as I previously stated, I think that for a new user
> searching for information on AMQP libraries, the obvious list to look
> for would be "amqp-libraries".  The name proton is pretty meaningless
> unless you already know it. (There is also perhaps some general value
> for us in having the association between "qpid", "amqp" and "library"
> become more obvious to search engines).

If you look at what "amqp library" pulls up now, it's really difficult
to see how a mailing list name will make *any* difference to the
results. What we really need is a highly connected web page that makes
mention of those words.

I also think that particular concern is only relevant for people who
already know about amqp, and we actually care about a much broader
audience than that. I think we need words like "simple, easy,
request/response, pub/sub, messaging, ..." to lead people to us as well,
and we're obviously not going to cram all of those into a mailing list
name.

> In any case I'm not sure that "stream of work" is the right way to
> define mailing lists for the user community. Stream of work is a
> concept that is meaningful to us, the committers, but a user will be
> looking for functionality - which is a more natural definition of a
> mailing list. The way that we choose to divide and sub-divide our work
> would seem subject to change, what is part of the current "Proton"
> stream may not always be so in the future.

So to be clear when I talk about "stream of work" I'm not talking about
a few people allocated to spend a couple man months on feature X. I'm
talking about an ongoing stream of work that will involve a multi year
dialog between the developers of said work and its users, and I think
the whole point of mailing lists is to facilitate that dialog, so the
stream of work is really all that's relevant.

I think generally when people join a mailing list it's because they want
to talk to a group of people, usually because they have a specific
problem they need help with, not because they want to discuss an
abstract category of functionality.

--Rafael



---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to