Bugs in Advogato

Posted 20 Jun 2007 at 22:31 UTC by zbowling Share This

The trust metric system seems to be chalk full of bugs lately. In relation to what bwtaylor said earlier, the trust metric system seems not to be working correctly. Other issues have been happening as well.

I've been testing on user "mentifex" after his crank/attack on the community, and I'm not able to mark him as an observer. While marking him as an apprentice shows up to me, it does not seem to stick.

Another bug when I try to syndicate my external feed, it does not seem to do anything.

There are other odd issues still lurking around the site as well. Anyone know what is going on with Advogato these days?

Bugs, posted 20 Jun 2007 at 23:21 UTC by robogato » (Master)

Can you clarify what you mean by your cert not "sticking"? The normal effect of rating someone as an observer (their default or uncert'ed state) is to remove an existing cert. So if you had previously certified mentifex as an apprentice, rating him as an observer would have the effect of removing your apprentice certification of him. You might want to check out the FAQ, which addresses this and a few other common questions about how the trust metrics work.

With regard to your feed, the most likely reason for it failing to show up is that the XML is invalid or not well-formed. Try validating your feed URL through a feed validator. If it validates, then it may be using a syndication format that isn't supported yet by mod_virgule.

You can find out what's going on with Advogato and mod_virgule in general by following the status reports in my blog.

There are still plenty of bugs to be fixed and features to be added. See the ToDo list for details. Items that are crossed off have been done. If you've got a bug or feature request that isn't on the list, let me know and I'll add it. Patches welcome.

Observer, posted 20 Jun 2007 at 23:23 UTC by mbrubeck » (Journeyer)

From the comments on the earlier article you referenced:

"Cert'ing as observer is the same as no cert, it's how you remove an existing cert if you've given them one, so the site hasn't been hacked (relax, take a deep breath :-)"

Your other complaints are pretty vague. How about posting journal entries with proper bug reports (steps to reproduce, detailed symptoms), and pointing an admin at them? Or you can look at the mod_virgule code yourself...

certs, posted 26 Jun 2007 at 13:07 UTC by lkcl » (Master)

there's no bug, zbowling.

if you check the degrees of separation between you and the nearest other 'Master' and upwards, you'll probably find that it's quite a long way between you and the top-level seeds.

so the 'flow', which isn't a floating-point number thing it's a fixed-point allocation system, alters constantly at the "ends", as other people temporarily get 'flow' instead of you.

what's supposed to happen is that every now and then, raph (or, actually, now it's stephen) is supposed to count the number of people on advogato and increment the capacity a little bit, to help avoid _some_ of the fluctuations.

but, essentially, you need to find someone who is a bit closer to the top-level seeds and ask them to Cert you.

the more people there are that are closer to the top-level seeds, the less fluctuation there will be of the whole network.

regarding mentifex, posted 26 Jun 2007 at 13:10 UTC by lkcl » (Master)


regarding mentifex: you need to be a little more tolerant of people who are clearly so far out there that it will be several years, possibly decades, before their work and motivation is understood.

there are many people who are autistic, to some degree, in the free software community and you are damn lucky to have them and, if your attitude is prevalent, damn lucky not to be one of them.

Fixing The Deviation?, posted 28 Jun 2007 at 05:10 UTC by nymia » (Master)

OK, someone will admit the deviation can be discomforting. Is there a way of moving beyong this point and see how that deviation remedied.

http://advogato.org/trust-metric.html, posted 29 Jun 2007 at 13:54 UTC by lkcl » (Master)

Assignment of capacities

The next step is to assign a capacity to each node in the graph. This is done by breath-first searching the graph from the seed, computing a shortest distance from the seed to each node.

Then, capacities are assigned simply as a function of this distance. Nodes closer to the root have high capacity, which diminishes with distance. Currently:

cap(0) = 800 cap(1) = 200 cap(2) = 200 cap(3) = 50 cap(4) = 12 cap(5) = 4 cap(6) = 2 cap(i) = 1 for i > 6

so, by the time you get to 6 degrees or more, very little "flow" comes your way. it still happens - just not as much.

solution 1: increase capacity (requires steven to recalculate and reassign)

solution 2: find someone closer to degree 0 (root-certs) to cert you.

current caps, posted 29 Jun 2007 at 14:49 UTC by robogato » (Master)

The trust metric caps were increased since the original description back in 1999. These are the current caps (in place at the time I took over from Raph):

cap(0) = 3200
cap(1) = 800
cap(2) = 200
cap(3) = 50
cap(4) = 12
cap(5) = 4
cap(6) = 2
cap(7+) = 1

As described in my recent blog entry, our problem lately is not so much with the capacity assignments but the lack of trust from seed users. Federico removed all his outbound certs a while back, taking a huge amount of trust out of the flow and leaving only raph as an active seed user (alan and miguel haven't logged in for quite some time). Fixing this is a matter of swapping the inactive seeds with new, active seeds who can inject a little more trust into the flow.

Minor bug, posted 1 Jul 2007 at 11:15 UTC by salmoni » (Master)

One trivial bug I noticed (not related to the trust metric sorry): Articles have text at the bottom with 2 hyperlinks: "Read more... (7 replies) (2 new)". Clicking on the "2 new" part goes to the last read comment (in this case, the last 3 comments) rather than the first unread comment (in this case the last 2 comments). I'm not sure if this is intentional (to show the reader where the discussion reached), but I don't think so.

Like I said, trivial!

Last read vs First unread, posted 1 Aug 2007 at 23:08 UTC by robogato » (Master)

Sorry, I only just noticed your comment about the message pointers, salmoni!

I took a look at the code and it appears to be intentionally taking you to the last read comment rather than the first unread comment.

I merged the message pointer feature into my codebase as-is from mod_virgule classic. The old CVS repository says it was merged into the classic branch in 2002 by gary from the codebase used for Advogato's long lost opposite, Badvogato. I don't know who wrote the original patch or what their intention was. For some reason, I have this vague memory that it might have been based on message pointer behavior in the old citadel BBS software.

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!

Share this page