News for hackers. Stuff that matters.
Re:Open Sourcing a Failed Startup by buro9 on Friday November 21, 2014 @12:12PM (#8642153)
I really feel for nirvdrum here, we had exactly the same hard process with Microcosm.

We were somewhat better placed to open it up, we had separated various parts more completely, hadn't put in the full billing engine, and had no third party code or other things that we were distributing that might be an issue.

But... one of the things nirvdrum didn't touch upon was that we just hadn't designed it to be a single-server, single-customer install.

We designed it as a platform, and made assumptions along the way. I didn't mind opening the embarrassing code (everyone's shit smells), but the people who were asking all seemed to want a single, well-documented install... and we were having to walk away and go get a day job.

We did choose to open it in the end, and you'll find our work here: https://github.com/microcosm-cc/

But I only know of a few people who have actually installed it to date (and struggled, because it's not well documented). All of the effort to open-source it just found another set of tyre kickers.

Sigh. It's tough. Just letting go. And it's definitely made tougher by wrangling with these questions and trying to do the right thing without investing so much at a time when you're exhausted, broke and likely quite down about things.

Re:Open Sourcing a Failed Startup by nirvdrum on Friday November 21, 2014 @01:06PM (#8642435)
Thanks. That's another excellent point. I think I had it in my post outline and just forgot to put it in here. The Mogotest application is complex to set up. For each version of IE you want to test, you need a separate Windows installation with that version of IE. To test Safari and iOS requires provisioning Macs (we used leased minis running VMs on ESXi). And the code itself runs on a Linux cluster.

Since our cluster provisioning and deployment was encoded in the project rubber [1] configuration, that complication is somewhat mitigated. But the point remains: the application requires multiple machines across platforms to be remotely useful.

[1] -- https://rubber.io/

Re:Open Sourcing a Failed Startup by gfwilliams on Friday November 21, 2014 @01:12PM (#8642467)
As one of buro9's users, I have to say I was very happy to hear that the software was going to be Open Sourced, however I was dreading having to migrate it all to my server.

In the end I'm much happier that buro9 found a way to keep it ticking over. Given the choice of continued service vs. Open Source I definitely would have chosen continued service - but it's even better to have both!

Re:Open Sourcing a Failed Startup by buro9 on Saturday November 22, 2014 @05:20AM (#8645264)
> the end I'm much happier that buro9 found a way to keep it ticking over

It's worth noting for people that don't know, that I ate my own dogfood. The communities and forums I run are now hosted on Microcosm.

So when we were facing the "What next?" moment, I knew I had to install and setup my own instance of the platform just to give my users an uninterrupted service.

What I did, was for every customer already on the platform who had stuck with us, loved what we'd done... I offered them space on my instance.

I never wanted to leave even a single customer in the lurch. I think it sucks that startups do, especially if a way can be found.

Re:Open Sourcing a Failed Startup by shurcooL on Saturday November 22, 2014 @04:07AM (#8645167)
Wow, I didn't know that... Your bluemonday package is one of higher quality and useful Go libraries I know.
Re:Open Sourcing a Failed Startup by buro9 on Saturday November 22, 2014 @05:17AM (#8645258)
That was one of the bits we had always intended to open-source. There was a real need for decent HTML sanitisation in Go.

When I find the time and energy (between the new job and getting married in a month), I still want to add CSS sanitisation to the mix as well.