- Open Source News Forums
- > iMatix to drop OpenAMQ support by 2011
- > Re: So what went wrong? Why other Big open source projects thrive?
Re: So what went wrong? Why other Big open source projects thrive? 02 April 2010 13:42
It is inaccurate to cite implementation and use of AMQP/0.9.1 (!) as
examples of participation in the AMQP development process.
Again, where is the stack of protocols? Where is the community of
protocol developers? Why is it so difficult to join the process? Why
did my team (of highly skilled engineers, such as Martin Sustrik and
Mato Lucina) abandon AMQP as a wreck, after the New York
face-to-face? They were involved in AMQP from the start, and yet
walked away, disgusted. That was in 2007. I remained involved in
the working group, hoping there would be some change in approach.
I recall at that meeting a formal vote where the working group agreed
to construct AMQP as a stack of protocols. Everyone agreed, except
JPMC and RHAT. That vote was ignored, mainly by yourself, as
chairman of the WG.
To say that I'm disappointed with the way the working group was
organized by JPMC and RHAT, from the start, is putting it mildly.
From the start JPMC allowed RHAT to force AMQP into incompatible
directions, such as the infamous 0.9 spec with its "work in progress"
sections. That was your work, John, as chairman and it cost us two
years of progress.
It is a shame that so much of what happened was private, hidden from
view. That does make it harder to discuss afterwards.
Even just forcing the AMQP process into public view took us years.
That effort should not have been necessary! How can you claim on the
one hand to be aiming at an IETF standard, and otoh maintain secrecy
and barriers to participation? It is not coherent.
Yes, Rabbit implemented a great product that interoperated almost
immediately. That proved the quality of AMQP/0.6 as originally used.
Today, it's impossible for anyone to build AMQP/1.0 code. I applaud
the Rabbit guys for their work and their success so far. RabbitMQ is
great! But excellent software does not translate into a successful
protocol, nor does AMQP/0.9.1 prove that AMQP is alive.
As for lock-in... AMQP/0.8, AMQP/0.9, AMQP/0.9.1, AMQP/1.0,... each
incompatible, each vendor with a different API, different management
tools,... AMQP is not compatible even with itself. Every successful
AMQP deployment is on AMQP/0.9.1, which is a dead, unmaintained spec
with no future. AMQP/1.0 has been delayed year after year. It
should have been out in 2007. This should be ringing huge alarm
bells. What firm would bet their business on such an opaque and
iMatix does have an obligation to its clients, which is to make their
systems work, reliable, and economically. We cannot in good faith
recommend AMQP, it's too costly, inefficient, and destined in our
view to become a "telco" standard, massive and expensive and used
only by the large and rich.
And 0MQ is a protocol project too. We will, carefully and using
fully open processes like Digistan's COSS, develop open wire
protocols that let anyone build interoperating applications. Take a
look at wiki.amqp.org, where iMatix is the only vendor writing
interoberability specs, and tell me again that we has an agenda to
lock in users.
31 Mar 2010 19:22Answerguy
31 Mar 2010 21:34Pieter Hintjens
01 Apr 2010 15:09Answerguy
02 Apr 2010 03:58John OHara
02 Apr 2010 13:42Pieter HintjensRe: So what went wrong? Why other Big open source projects thrive?
- Threaded View
- Flat View