Jun 16

Sharing the load effectively

At the end of my last post I said I was wandering off to think about scalable, low-overhead recommendation systems.

It’s funny how preconceptions work. I know, I think better than most people, how often decentralizing systems to avoid single points of failure is good engineering. Yet I had to really struggle with myself to jettison the habits of thought that said “If you want to use money to help people, you’re going to have to build a centralized, heavyweight structure around the management of that task.”

But struggle I did. Because I’d already tried that, and failed.

I also had to get past the idea that identifying good funding targets can be crowdsourced. Nope. Identifying candidates and digging up information on them can be, but actually evaluating merit and centrality will take knowledge most contributors not only lack but have no strong reason to try to acquire.

Once I got my head clear, this is what came out:

http://www.catb.org/esr/loadsharers

The basic trick here is piggybacking not just on the payment transfer capacity of remittance systems but on their patron/client communications channels as well. That way Loadsharers doesn’t need to manage anything itself other a handful of adviser web pages and a bunch of trust relationships.

Also notice the implications of how I designed the Adviser role. By the time we have a half-dozen or so advisers I won’t be key man anymore. That’s intentional.

I also like the fact that there will, in effect, be a (mildly) competitive market in adviser skill, with loadsharer contributors tending to gravitate to advisers who exhibit activity and diligence. That’s intentional too.

Jun 15

Load-Bearing Internet People

I just finished giving a talk – by remote video – at South East Linux Fest, about the Load-Bearing Internet Person problem.

An LBIP is a person who maintains the software for a critical Internet service or library, and has to do it without organizational support or a budget backing him up.

That second part is key. Some maintainers for critical software operate from a niche at a university or a government agency that supports their effort. There might be a few who are independently wealthy. Those people aren’t LBIPs, because the kind of load I’m talking about isn’t technical challenge. It’s the stress of knowing that you are it and you are alone, the world out there has no idea what a crapstorm it would be if you failed at your self-imposed duty, and goddammit why doesn’t anybody care?

LBIPs happen because some of the most critical services can’t be monetized. How do you put a meter on DNS? Or time synchronization? Or having a set of ubiquitous and reliable crypto libraries? Where there’s no profit stream, markets are not going to directly solve this problem.

I know at least two LBIPs whose health has broken under that strain – Dave Taht and Harlan Stenn. Me, I’m still generally healthy, but my recent medical issues have re-focused my mind on the LBIP problem.

I spent seven years trying to solve this problem by founding an organization to collect funds from sponsors and distribute them to LBIPs. That was the Internet Civil Engineering Institute. It shut down late last month because, as it turns out, recruiting people who are both willing and entirely competent to run an organization like that is really difficult. I failed at it.

(I’ve designed and founded two nonprofits that survived my departure and are still on mission, one 17 years on and the other 26 years on. I’m actually good at that game, but ICEI failed anyway. Possibly someone smarter or more streetwise than me could have made ICEI work, but given my previous track record of success I don’t think that would be a smart bet.)

What I said at SELF was this: centralized attacks on the LBIP problem have failed, so we need a decentralized, distributed one. Services like Patreon, recurring PayPal remittences, and SubscribeStar give us the technology to do that. What we need to add is consciousness about the problem and some social engineering.

Here’s the challenge I put to the audience there: If you have a good paying job, earmark $30 a month – the equivalent of a moderately-priced restaurant meal. Identify three LBIPs. Remit them $10 a month.

Then go to every gainfully employed programmer you know and explain to them why they should do the same thing, and also further spread the word.

The fanout is important. One of the failure modes we want to avoid is for all that support to go to a handful of highly visible hackers like, er, me. There are lots of LBIPs working in obscurity; we need to solve this problem at scale, not just for a few prominent figures.

The SELF audience liked this idea – and then somebody raised the question I should have expected: “How can we know who to fund?”

Sorely tempted as I was to say “There’s always me…”, I didn’t. That would have been a humorous answer of the funny-because-it’s-true kind, but the discovery problem is a serious one. Several other questioners chewed on various possibilities. I ended up saying I would try to jump-start a discovery process on my blog by collecting a list of LBIPs.

That’s not going to be a solution that scales well, though. We’ll have to feel our way to a better one; I have some ideas which I’ll develop in future posts.

I do have a name for the effort – thought it up a few minutes ago. Loadsharers. We need to work out how to be effective loadsharers.

For now, my comments are open. Please check in if you (a) want to take the loadsharer pledge – $30 in 2019 dollars to one, two, or three LBIPs every month (ideally three), or (b) have an LBIP to recommend.

I will curate a list of LBIPs I think are worthy. I should not be the only person doing this. Eventually we’ll set up a recommender system and a way for LBIPs to declare funding goals. Mumble web of trust? Something like that should be doable.

Please do not wander off into trying to design a better mediation/discovery system in this comment thread (yeah, I know my audience). Save that for my post on that topic, coming soon.

As final and obvious point: yes, I think I’m a worthy LBIP, go ahead and do that $10 thing at me, initially. (Note to self: create a “Loadsharers” tier.) But I have a relatively low monthly figure that I consider “enough”; above that, I’d really rather the money went to other people.

So don’t be surprised if, a few weeks down the road, you get a patron notice from me saying “Enough! Roll a D6 and if it comes up 5 or 6, drop me and go fund someone else.”

/me wanders off to think about scalable, low-overhead recommendation systems…