Not Having Any Fun

Ok, so I mentioned I was working on a new thing that involved moderation, administration and the auth system I put together as part of the Four-and-a-half-and-counting part series on Authentication. I've still got one or two left to write there, but since this "don't talk about it 'till it's done" thing worked out so well, I'm going to keep you in suspense.

The result of my toil is Nitrochan a massively-ish scalable, real-time message board system inspired by the *abas that the internet is so full of. My problem with 4chan and similar boards is that they are sort of like going to a restaurant and having a guy come by to shit on your plate every few minutes. It seems that what you'd really want[1] is a constant, flowing stream of shit that you can pan for nuggets at your leisure. And this is an attempt at that. When a new thread is started, the boards are all updated with new data. When a new message is posted, the appropriate threads move up the sort order, and people already on the thread get the new message via Comet rather than having to F5. Threads can be moderated and moved between boards through similarly soft-real-time mechanisms.

The github is there, released under the terms of the AGPL[2]. I'll have another go at setting up an instance here for my own nefarious purposes[3] later this week.

The UI layer is still somewhat incomplete for a message board; we can't designate images as spoilers/nsfw, there aren't any comment markup options yet, there's no way to proactively protect a board or thread from spam, and the RSA login process is just as manual and painful as it was the last time I discussed it.

Still, we've got a good starting point to look at in terms of putting a running system together[4].

Now then, the bad stuff.

Bad Stuff

The Erlang deployment process is really beginning to annoy the fuck out of me.

I mean, it kind of did last time too, but I figured that it would get simpler as I went on and automated pieces. That... didn't really happen. You'll note that I mentioned I'll be trying again to set up an instance of Nitrochan.

The attempt proved to be futile, even without having to wrestle with rebar again. I'm really beginning to grudge that the language designers seemed to have considered actual deployment of an app to be outside of their scope. That's a shame, because every useful application is going to need to be deployed somewhere, and doing this stuff manually gets really tedious if you rely on even two or three libraries not found in the core Erlang image. rebar would be a good solution, from what I understand about it, assuming it did what it says on the tin. It has yet to for me.

That's saying nothing of the massive headaches I've gone through as a result of platform incompatibilities. Basically, I spent about an hour trying to figure out why it only works on my machine, only to remember that my deployment environment is a 32-bit Debian machine while my development environment is a 64-bit build of the same. That's really the only thing that could possibly make a difference, because I've painstakingly reproduced my local directory structure, installed programs and downloaded libraries, but running the same make tasks on the same code seems to crash on the server even while working perfectly well on my local. This is Not Inspiring Confidence©™ in Erlangs' touted cross-platform abilities. Because if I can't depend on my program running the same way on two builds of the same OS on different architectures, how am I supposed to believe that it'll do anything but explode when I try to deploy on another OS?

Bottom line, I'm not having any fun.

I did learn a lot about concurrency outside of the lock/mutex world, and I appreciated the opportunity to mess around with actors on a grander scale than I would usually be permitted, but the continuing headaches aren't worth it for me so far. I may come back to it once I've recharged my mental batteries. For the next week or so, I'll be playing around with Clojure[5].

---

Footnotes

1 - [back] - If that was your thing.

2 - [back] - Read it before hacking on it, and get my permission if you want to use the system under another license.

3 - [back] - Really, it's just so that the Toronto Lisp User Group can have something better than mailing list with which to communicate between meetings.

4 - [back] - And one of the items on that roadmap is automating the RSA login process.

5 - [back] - Links added to the sidebar, and shiny new logo in the bar up top.