Is anyone here interested in starting work on a similar (but new) project?

[Mark Taylor]

Hi,

Is anyone here interested in starting work on a similar (but new) project?

When I first came here a year ago, I had already started on a similar project of my own, and I've made a lot of progress, and hope to have a written spec finished within the next 2-3 months.  It would be interesting to bounce ideas off other like-minded people, and even better share the work!

regards, Mark.

[Daniel Kersten]

Hi Mark,

My own dataflow project got derailed as I didn't have enough time to work on it properly. I still often think about it and jot down notes, but very little actual work is getting done on it, sadly.

I'm interested but don't know what kind of time commitment I could make (I'm cofounding a startup - nothing related to dataflow programming though), but feel free to email me directly to discuss ideas. I'm always interested in chatting about dataflow ideas and don't have enough people to do that with.

Regards,
Dan.

[Duane Johnson]

I think I failed to post this to the list.

---------- Forwarded message ----------
From: "Duane Johnson" <duane....@gmail.com>
Date: Jun 7, 2012 11:55 AM
Subject: Re: [anic] Is anyone here interested in starting work on a similar (but new) project?
To: <dker...@gmail.com>

Hey I'm cofounding a startup too! I would be happy to be a test user, and someone to bounce ideas off of, but I probably would not be able to contribute to the programming side.

[Daniel Kersten]

Oh yeah, as Duane said, I'd also be happy to test your language.

[Daniel Kersten]

Hey everyone,

So ANI was reposted Hacker News today and it caused a flurry of activity around ANI before people started realising that the project is dead.

So I was thinking, perhaps it is time to revive the ideas behind ANI in a dataflow language project  in the spirit of ANI, based on the ideas and concepts of ANI, dataflow, stream processing and signal processing, but with a much scaled-down scope and less ambitious goals.

So, Mark, if you would like to share your work, now may be a good time to do so! Otherwise, is anybody else interested in such a project?

I would suggest that we work out the semantics first of all. Something similar to ANI maybe. Then we rework the syntax - the ANI syntax seems to have been a bit of a friction point when people are introduced to it. Perhaps we could base the syntax on Python or Ruby, to some extent? Or if Mark wishes to share his work, I assume a lot of this will be done already?

After that is done, I think the goal should be to get something running ASAP so that people can actually write and test code. This means no native code compilation, no faster than C, no fancy complex runtime, no optimising compiler. Maybe a really simple high level bytecode that the source gets compiled to and a very simple virtual machine/interpreter to run it.

I'm not sure what kind of time commitment I can make (startup and all that), but dataflow programming has been on my mind for a number of years, so its something I'm willing to put a few hours into every so often for the foreseeable future.

So, what do you guys think? Let me know!

Regards,

Dan.

[Mark Taylor]

Hi Dan,

Your previous post suggested that you couldn't put any time or effort into the project, but this latest post seems to say different!

I am at a fairly advanced stage with my language, but when I previously said I was 2-3 months away from starting a written spec, that was maybe a bit ambitious.  I've taken a bit of a break over the last few weeks, but I can feel a new wave of energy coming on soon.

I am happy to discus ideas, or equally happy to get something down in writing for people to read (but this will be at least 3-4 months at a minimum).

When I looked at the syntax of ANI, I was a bit repelled, and so I designed a syntax that is much simpler (but hopefully just as powerful).  Where ANI has a "horizontal" syntax like module1 -> module2, I have vertical implied pipes with no explicit pipe syntactical sugar. e.g:

  module1

  module2

where the pipe is implied vertically - you can think of it working like this:

module1

 |

V

module2

I also came to the conclusion that a new data-typing and object-oriented model is necessary, which has added work and delayed my written spec somewhat.  I want to have generic black-box modules of code processing streams of data objects, therefore the code & object-model has got to be fairly clever if strong & secure data-typing is going to enforced.

One of the ideas I am playing with is that code will be dynamically compiled at runtime - this will provide several big advantages.

regards, Mark.

[Rob Bird]

(caveat: I was the founder and CTO of this company at one time)

Red Lambda's MetaGrid exposes a dataflow programming model mapped over a P2P network. It's not available for application developers to my knowledge (maybe if you contact them). My point is that the concept works, and is definitely well-suited to stream processing at scale because the dataflow model maps naturally to a distributed environment.