[openamq-dev] Spec conformance chart? Shopping for a new broker

[Pieter Hintjens]

On Sat, May 10, 2008 at 12:18 AM, Brad Clements <bkc at murkworks.com> wrote:

> Ouch. So to transfer files through OpenAMQ, I'll just stuff their
> contents into a single gigantic "basic" message type!?

There are several levels at which semantics like file transfer can be expressed.

You could indeed stuff files into huge messages but that breaks the
whole point of staging (for recovery if things break).

Look at http://wiki.openamq.org/package:simple-file-transfer, you'll
see an example of how to do file transfer by sending chunks across as
separate basic messages.

> Or has the WG members decided that file transfer should be entirely out
> of band and are backing away from supporting the concept of "really
> large message" transfer within the amqp spec?

It's not either-or.  The use of content classes is/was contentious
since they combines functionality and properties.  I like that because
it lets us add packages of functionality to the network in a clean
fashion.  Other people dislike it because it breaks the unitary model
of how messages are transferred, and a unitary messaging model is
necessary for certain kinds of work (consider: are file transfers
covered by transactions, and if s, how).

In fact, it is quite straight forward to implement higher level
semantics like file transfer on top of AMQP.  You can imagine a file
transfer 'broker' that implements staging semantics, acting as an AMQP
application, and talking to file transfer clients over AMQP.  The
server itself does not need to know anything about these semantics.

-Pieter