Advo Gato

A combination of advocacy and avocado. Experimental hypertext, like and unlike WikiWikiWeb. A social event for FreeSoftware hackers.

Wiki allows anyone to edit anything, and doesn't require logins.

Advogato goes in the opposite direction: people are inserted into a fairly strict and rigid certification scheme. We have yet to see whether this works well or not. Therefore, like Wiki, it provides an interesting experiment in how collaborative spaces work.

Could someone summarize their TrustMetric please?

At the moment (April 2000), they have four levels: Master, Journeyer, Apprentice, and Observer. There is an active debate about the difficulty of squishing all the things you know about somebody into two bits of information: e.g. people might be more or less active contributors; agree with your political position or differ; be a good programmer or bad; they might write English well or not; be good judges of others or not; and so on.

[Hmm, I would suggest that Advagato choose what its terms mean, and then stick to a strict, one-dimensional view. For example, the ranking can exclusively indicate how much the user has "mastery over the machine" and that is all. That would require the top seeds to be masters, however, but that's probably not too hard. Other ideas, such as active contributors, etc, could be better handled as the "trust tower" at OhlOh?.net. People will have to define their site based on their primary trust index or axis.]

There was briefly a Dimwit level as well. This was removed when users began to use it for spite rather than for humor as intended.

Broadly: after you sign up, people who know you assign you to a level. The strength of this accreditation depends on their own level and certifications, through a GraphTheory algorithm. You can only certify up to your own level. The graph is seeded with four famous people.

One AdvoGato member has decided to become a BlackBox (aka TitForTat) user who certifies other users exactly as they certify HimOrHer. This experiment in SocialEngineering is intended to produce effects in both the social and technical fabric of the TrustMetric.

Which member is this? "Ah, that would be telling."

The nature of AdvoGato is for users to certify other users based on their perceived competence, not just as a BlackBox. It will be very interesting to see if the TrustMetric is able to ignore this DeviationFromThePlan?.

Well, I'm *obviously* a master. Other people's ability to discern this fact is the best measure of their competence. The BlackBox strategy works fairly well in this case. -- an anonymous master (?) having a bit of fun

To me, the really interesting thing about the AdvogatoTrustMetric is that it's designed from the beginning to be robust. Even if the certifications are somewhat 'noisy', or if unwashed hordes of outsiders are trying to get in, it will be rather conservative in letting large numbers of people in. Yet, it is designed to be very fast and nimble in terms of letting in people who actually match the criteria. A large part of why I built Advogato was to test this theory in practice. Guess what? It's working pretty well.

One of the great challenges in building DistributedComputing systems is to come up with new, more flexible forms of authentication. The old password and AccessControlList technique isn't very secure, and doesn't scale well. The success of Advogato suggests a new approach: don't let people do arbitrary amounts of damage/good work once they're authorized. Instead, limit the amount of change allowed by any user, and use CapacityConstrainedFlowNetworks to set this limit.

A lot of the bad DistributedComputing design I see these days is due to overly inflexible and restrictive authentication mechanisms. One of the great things about the Wiki experiment is that it shows the immense power even a simple system can have when you remove these restrictions. Abuse is always an issue, but Wiki somehow seems to have come to terms with it. But I'm afraid the 'open to everyone' policy just won't scale to systems where you actually have money riding on the ability to make changes. -- RaphLevien

In other words, it just won't scale, period.

Also consider the difference between mistrust and distrust. Here is a simple proof that it exists:

There is a perfectly reasonable and trustworthy text who a certain individual believes was written by a hated and excluded user or at least a HatedIPRange. Thus no one is allowed to read about this concept, nor to copy "unacceptable" amounts of text from it. Numerous methods of intimidation are used, and those who do not demonstrate the RequiredMistrust? of the HatedIPRange have their own text targeted.

By failing to note the difference between distrust (an organized separation of powers, enforcement of a due process) and mistrust, one is rather doomed to end up empowering the bigots and others who simply don't care about the final results at all: social clubs take over.

The now defunct is much more the right idea, as it seems likely to do what reputation usually does, which is, label and exclude. There is no such thing as a "good reputation" anyway - it's just a means of fraud, especially if it can be traded from one to another. And, "bad reputation" opens the door to various manipulations, such as promoting ideas one wishes to extinct under the "bad" name or range, thus creating a field of hate and intimidation around them as the usual suspects react to the name.

20 October 2005 google search for advogato yielded 933,000 results and 1 sponsored link, "Google is hiring expert software designers!"

Some advogatans on WikiWikiWeb:

View edit of June 4, 2013 or FindPage with title or text search