Enterprise-Ruby Wish List by Francis Cianfrocca

Z

zoat

In all the recent talk (some would say hype) about the Ruby programming
language,relatively little has been said about Ruby's usefulness in
enterprise development shops, but that is beginning to change. Francis
sidesteps the usual debate by turning the question around. What do
enterprise developers need, that they're not getting from their tools
today? Based on the answers to that question, he looks at what Ruby
currently has to offer in the area of entrprise infrastructure.
http://www.infoq.com/articles/Enterprise-Ruby-Wish-List
 
J

Joel VanderWerf

zoat said:
In all the recent talk (some would say hype) about the Ruby programming
language,relatively little has been said about Ruby's usefulness in
enterprise development shops, but that is beginning to change. Francis
sidesteps the usual debate by turning the question around. What do
enterprise developers need, that they're not getting from their tools
today? Based on the answers to that question, he looks at what Ruby
currently has to offer in the area of entrprise infrastructure.
http://www.infoq.com/articles/Enterprise-Ruby-Wish-List


Why no mention of http://www.pragmaticprogrammer.com/titles/fr_eir/ ?
 
J

Joel VanderWerf

Francis, you make good points in your reply (snipped). I was asking why
the blurb claimed that "relatively little has been said about Ruby's
usefulness in enterprise development shops", though there is a book on
the subject. I should have made clearer I was talking about the
ruby-talk posting, and not your article.
 
J

Jeremy Tregunna

I'm guessing that you're asking why I didn't mention that book (or any
other) in my piece. The short answer is that the piece is a survey of
existing infrastructure, not a literature survey. A slightly longer
answer
is that I was trying to address needs that are faced by enterprise
developers *in general,* not needs faced specifically by Ruby
developers. I
think that enterprise software infrastructure can benefit from some
new
components written in Ruby, *irrespective* of whether the development
language you subsequently choose for any given project is Java,
Ruby, or
something else. As always, "use the best tool for the job," and I'm
trying
to explore whether some new enterprise-class infrastructure can,
and should,
be written in Ruby rather than the default choice (Java).

A small example: let's say you're a Java programmer accustomed to
using JMS.
If I offered you a message-queueing product that supported JMS but was
"better" in measurable and compelling ways than what you use now (and
parenthetically happened to have been written in Ruby), wouldn't
you want to
give it a try?

Not necessarily. I mean, I have Ruby experience, but if I didn't, I
don't know as that I'd want to have to learn a new language to use a
new tool, even if it is "better" (unless it's vastly "better" and not
just marginally).
 
A

ara.t.howard

A small example: let's say you're a Java programmer accustomed to using JMS.
If I offered you a message-queueing product that supported JMS but was
"better" in measurable and compelling ways than what you use now (and
parenthetically happened to have been written in Ruby), wouldn't you want to
give it a try?

well, i heard you can catch java - that'd scare me off.


;-)

-a
 
R

Richard Conroy

Fair enough. But if you were the Java programmer in this example, would you
be afraid you might catch ruby?
I wouldn't mind being the cause of that infection ;-)

Well as one of those java programmers who is really seriously looking at
using Ruby for enterprise stuff, I am going to chime in here.

My initial opinion: Ruby is a really sweet sell to experienced Java developers.
Its conceptually very easy to get productive quickly with it, and
'Least Surprise'
is not blogger speak, but an actual real phenomenon. When half of the
recursion you have written in your professional career is in your first non-toy
ruby program it shouldn't work straight after you iron out the last interpreter
error.

In fact any Java developer who has been with it from the start is in a really
good position to evaluate Ruby with an open mind, as many of the topics being
discussed about Ruby/Rails prime time readiness, are things that Java
developers have lived through.

Mostly enterprise developers are most interested in 'can do anything'. Which
more or less translates to:
- runs on anything
- talks to any system
- has good libraries

Certainly as an enterprise developer I have seen projects get
rewritten when a library
comes up short and can't do the job. Its an area of concern for me that the
enterprise libraries in Ruby have 'Wet Paint' signs on them.
Enterprise developers
expect the libraries to be there, and to just work as advertised. This
translates
to being complete, and proven in the field.

Depending on what kind of enterprise work you are doing, you are generally
not that interested in performance if you are honest with yourself. Certainly
a lot of the performance issues I see are due to poorly conceived application
protocols between machines.

Enterprise developers care about tools and support ecosystems (possibly
a bit too much), and can't understand why Ruby and Unicode generates
hundreds of heated arguments and very little quality technical discussion.
They mightn't even need it, but the fact that the debate is so heated is
worrying.

Theres also other little things that we worry about, like bundling Ruby
applications conveniently (doesn't seem hard, unless you are talking about
Rails), obfuscating source for support reasons (not IP-protection, but to
protect against meddlers) and how cross platform it really is (counting
everything here, including 3rd party library portability, performance parity
etc.) - I mean Rails for instance seems *very* unix biased.

Some other points - that Prag Prog book nailed it with 'Low Ceremony
Distributed Apps' btw. He totally described the enterprise app mentality
to distributed computing. I am warming to that book already - I have
climbed out the smoking ruins of two many big-M Middleware
solutions.
 
B

Brian McCallister

In my example, what I asked was: what if I offered you a message-
queueing
product *that supported JMS* ...... and happened to be written in
Ruby.
No one's asking you to learn a new language, that would be
singularly poor
marketing. ;-) My point with this example is that there are
elements of
today's application-support stack that could do with some major
improvements, and Ruby might be a good language to use for them.

How about one written in Java which supports Ruby (and C, Python,
PHP, Perl, C#, C++, Pike, as well as JMS)?

1) Install and start ActiveMQ ( http://incubator.apache.org/activemq/ )

2) 'gem install stomp' or http://svn.codehaus.org/stomp/ruby/trunk/

3)
--
require 'stomp'
client = Stomp::Client.new("username", "password", "localhost",
61613)
client.subscribe("/queue/SOMETHING.NIBBLE") do |msg|
puts msg.body
end
client.join
--
require 'stomp'
client = Stomp::Client.new("username", "password", "localhost",
61613)
10.times { client.send("/queue/SOMETHING.NIBBLE", "Hello!") }
--

4) See http://incubator.apache.org/activemq/stomp.html for more
details on getting the exact behavior you want.

5) See http://dev.tirsen.com/trac/activemessaging for a framework
around it which is designed to play nicely with a certain popular web
framework in Ruby.

6) See also Gozirra ( http://www.germane-software.com/software/Java/
Gozirra/ ) for a lighter-weight messaging solution which supports the
same spread of clients.

-Brian
 
A

ara.t.howard

Thanks, Brian. Actually I know a fair bit about ActiveMQ, and it's the main
reason why I think the Ruby world would do well to come up with a
competitor. You're making a classical case for not re-inventing the wheel,
but ActiveMQ is not the easiest product to use, nor is it particularly
graceful to require all manner of external bindings. It's also the kind of
committee-designed rather than use-case-driven effort you'd expect from
Java. Perhaps the obvious rejoinder is that a little pain never hurt anyone,
but then the success of Rails suggests that there may be a market for a
better way. There are other, more widely-used solutions in the Java world
that are commercial and very expensive. And there is also a very-well funded
project afoot in the Java world (AMQP) which is intended to bring real
standardization to message-broker implementations and reduce costs. That is
another strong clue that there is an unfilled gap. I've been working on a
Ruby implementation of AMQP, which should be ready for early looks quite
shortly now.

please release early! i've been looking for such a beast to base a
peer-to-peer clustering idea on...

-a
 
B

Brian McCallister

Thanks, Brian. Actually I know a fair bit about ActiveMQ, and it's
the main
reason why I think the Ruby world would do well to come up with a
competitor. You're making a classical case for not re-inventing the
wheel,
but ActiveMQ is not the easiest product to use, nor is it particularly
graceful to require all manner of external bindings. It's also the
kind of
committee-designed rather than use-case-driven effort you'd expect
from
Java.

How is it difficult to use?

curl -O http://.../incubator-activemq-4.0.2.tar.gz
tar -zxvf incubator-activemq-4.0.2.tar.gz
sh ./incubator-activemq-4.0.2/bin/activemq

At which point it is listening on port 61616 for its native protocol
and 61613 for stomp. I'll post a screen cast demonstrating it, if you
like.

Would you mind expanding on "committee-designed" I am very curious
about what makes you say so. James and Hiram account most of the core
design, though quite a few other folks have been contributing.

In terms of external bindings, you mean the fact that we specced out
the Stomp transport protocol independently of ActiveMQ? If you look
through the stomp-dev archives you'll see a lot of the same names as
on activemq-dev (plus a number of others). If I thought it added much
value I'll happily roll an activemq gem so the names match ;-)
Perhaps the obvious rejoinder is that a little pain never hurt anyone,
but then the success of Rails suggests that there may be a market
for a
better way. There are other, more widely-used solutions in the Java
world
that are commercial and very expensive. And there is also a very-
well funded
project afoot in the Java world (AMQP) which is intended to bring real
standardization to message-broker implementations and reduce costs.

Do you mean Qpid, nee Blaze, ( http://incubator.apache.org/projects/
qpid ) which is the code developed by JPMorgan, Redhat, and IONA
which is aiming to also join Apache? There are both Java and C++
based servers in the works, though the AMQP protocol is still
evolving behind closed doors :-(

AMQP is a very interesting protocol, it looks and feels, to me, more
like an API expressed in bytes over the wire than an app level
transport protocol because so much of the messaging server's internal
capabilities are specced out as part of the wire protocol.
Interestingly, ActiveMQ was designed by a small group of folks with
lots of commercial messaging server experience, AMQP and Qpid by a
committee of vendors with one big customer :) The Java version does
use MINA though, which rocks!
That is another strong clue that there is an unfilled gap. I've
been working on a
Ruby implementation of AMQP, which should be ready for early looks
quite
shortly now.

Sweet! Make sure to mention it on the qpid dev list, they'll be excited!
I don't know anything about Gozirra. Have you used it?

Only early versions for mucking about purposes. It is basically a
minimal Stomp server implementation.

-Brian
 
B

Brian McCallister

Any plans on adding ssl connectivity to the ruby stomp gem?

Certainly doable, I just have always used it on trusted networks :)
Will prod it this week coming week, time allowing.
And has it seen much production use? Hard to tell from the website.

The Stomp connector for ActiveMQ has seen a lot of use (and has grown
the knobs that accompany).

The ruby connector has fewer war stories, but has served me and a
couple others fairly well so far. Not sure the extent of its use. The
protocol is downright trivial, so adjusting a client is pretty
straightforward (I think).

-Brian
 
D

David Vallner

--------------enigC6365D70D34BAE1367528BE4
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

Francis said:
ActiveMQ is not the easiest product to use, nor is it particularly
graceful to require all manner of external bindings. It's also the kind= of
committee-designed rather than use-case-driven effort you'd expect from=

Apache software cumbersome and hard to use! Further in the news at nine:
Water is wet!

Sheesh.

The fact that specifically Apache developers have a penchant for writing
documentation on the edge of useless is rather known. (Witness the
Struts taglib docs. If you dare.) The fact more or less everything they
ship relies on two-three classes from commons-collections, requiring you
to download the horrendous iterators and stuff that was long since added
to the JDK from when that library came to be along, and commons-logging,
that I have yet to see do anything else than pointlessly wrap log4j or
JDK1.4 wrapping for the whole time an application exists also tells
books about what to expect in the ways of graceful dependencies.

And all that would be fine if you weren't provably wrong, since despite
all flaws, Apache projects on the whole are still able to knock up
foolproof startup scripts. (Thanks to Brian McCallister for pointing
that out.)

Also, the overgeneralising vaguely to nag at [insert name of language
that isn't Ruby here] doesn't quite make me trust you more as a
technology journalist. It comes across as so much cheap
community-pandering.

If you have nothing good to say...

David Vallner
(Can we -finally- stop with the cheap shots at not-Ruby?)


--------------enigC6365D70D34BAE1367528BE4
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: OpenPGP digital signature
Content-Disposition: attachment; filename="signature.asc"

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (MingW32)

iD8DBQFFJwlOy6MhrS8astoRAoLQAJ0Qs/8bQdn7qJJ9B2fU/Q6iWd+sVACffYjz
YjjSjf+UPruK2RXCjU2abC8=
=u+/B
-----END PGP SIGNATURE-----

--------------enigC6365D70D34BAE1367528BE4--
 
S

snacktime

The Stomp connector for ActiveMQ has seen a lot of use (and has grown
the knobs that accompany).

The ruby connector has fewer war stories, but has served me and a
couple others fairly well so far. Not sure the extent of its use. The
protocol is downright trivial, so adjusting a client is pretty
straightforward (I think).

Well, after playing around for only an hour I managed to run into this
bug in activemq when using stomp. Basically, if a consumer
disconnects in a bad way the broker can get stuck, and you'll never
see your messages.

http://issues.apache.org/activemq/browse/AMQ-724
 
B

Brian McCallister

Well, after playing around for only an hour I managed to run into this
bug in activemq when using stomp. Basically, if a consumer
disconnects in a bad way the broker can get stuck, and you'll never
see your messages.

http://issues.apache.org/activemq/browse/AMQ-724

Did you use 4.0.1 or the upcoming 4.0.2 (in RC and being voted on
right now)? As the bug mentioned, it's been fixed in 4.0.2.

The binaries being voted on:

http://people.apache.org/~chirino/incubator-activemq-4.0.2-RC4/maven1/
incubator-activemq/distributions/

-Brian
 
T

Tim Uckun

I took a brief look at activemessaging. Looks great. Then I take a
look at the roadmap and see that all the tickets logged are actually
spam.

No tickets logged, no milestones defined. I take a look at the
timeline and it's all spam too so no code checked in either.

To me this is one of the biggest frustrations of dealing with ruby and
rails. The libraries are in a unfinished state and are not being
worked on actively. Examples.

The DBI ADO library is b0rked. The answer seems to be to use ODBC. I
volunteered to work on the libraries if somebody could integrate my
work with DBI but nobody is interested.

I was excited about railscron and it's been dumped and replaced with
something which has no cron functionality at all (threading problems
from what I can gather).

I tried various AGI implementation for asterisk and the same story.
RAGI hasn't been touched in a year, the rubyforge forums questions go
unanswered, patches that were submitted are not integrated, bugs not
fixed, etc.

I don't know what the answer is. I am grateful that talented people
have dedicated their time and effort to create these libraries in the
first place but it's very hard to tell which libraries are robust,
which ones are abandoned, which ones are not being worked on because
they are stable and don't need any more work.
 
K

khaines

The DBI ADO library is b0rked. The answer seems to be to use ODBC. I
volunteered to work on the libraries if somebody could integrate my
work with DBI but nobody is interested.

I'm interested. Give me a good DBD::ADO and I'll get it added to DBI.


Kirk Haines
 
A

ara.t.howard

I took a brief look at activemessaging. Looks great. Then I take a
look at the roadmap and see that all the tickets logged are actually
spam.

No tickets logged, no milestones defined. I take a look at the
timeline and it's all spam too so no code checked in either.

To me this is one of the biggest frustrations of dealing with ruby and
rails. The libraries are in a unfinished state and are not being
worked on actively. Examples.

The DBI ADO library is b0rked. The answer seems to be to use ODBC. I
volunteered to work on the libraries if somebody could integrate my
work with DBI but nobody is interested.

I was excited about railscron and it's been dumped and replaced with
something which has no cron functionality at all (threading problems
from what I can gather).

I tried various AGI implementation for asterisk and the same story.
RAGI hasn't been touched in a year, the rubyforge forums questions go
unanswered, patches that were submitted are not integrated, bugs not
fixed, etc.

I don't know what the answer is. I am grateful that talented people
have dedicated their time and effort to create these libraries in the
first place but it's very hard to tell which libraries are robust,
which ones are abandoned, which ones are not being worked on because
they are stable and don't need any more work.

google 'rpa' (ruby production archive). for some reason, it didn't take off
like gems - imho it was a powerful idea...

cheers.

-a
 
S

snacktime

Did you use 4.0.1 or the upcoming 4.0.2 (in RC and being voted on
right now)? As the bug mentioned, it's been fixed in 4.0.2.

The binaries being voted on:

http://people.apache.org/~chirino/incubator-activemq-4.0.2-RC4/maven1/
incubator-activemq/distributions/

I was using 4.0.1. But 4.0.2 is even worse. In addition to the bug
still being there, activemq has a memory leak, at least on freebsd,
and I ran into another bug where it hangs on shutdown. Definitely not
ready for production use by my standards.

Chris
 
T

Tim Uckun

I googled RPA and all I got was a wiki filled with spam.

Here is a suggestion for people running wikis and trac.

Use captcha.
 
A

Austin Ziegler

I googled RPA and all I got was a wiki filled with spam.

Here is a suggestion for people running wikis and trac.

Use captcha.

Better yet: don't. Better to use a Wiki that does other proactive
stuff. Most CAPTCHA implementations wouldn't even remotely pass any
disability testing.

-austin
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

No members online now.

Forum statistics

Threads
473,744
Messages
2,569,484
Members
44,903
Latest member
orderPeak8CBDGummies

Latest Threads

Top