Social Networks and Advogato

Posted 29 Oct 2006 at 22:00 UTC by Ankh Share This

Like many others I'm delighted to see Advogato kept alive. But the recentlog page is much quieter than it used to be. Is this for social reasons, technical reasons or both? In planning "social network" features for a Web site I edit about old books I looked around at a number of other social network sites and concluded...

and concluded it's probably a bit of both.

Of course, a lot of people left when they thought that Advogato was going away. Others wanted some specific feature or other, such as better support for images or for threaded discussions. Some felt that they wanted to talk about things inappropriate for advogato.

Elsewhere, most of the successful sites have some sort of "goal", almost like a gaming score. Advogato has three levels for people; people on (say) deviantArt talk about pageviews (a meaningless statistic displayed on a user's profile, roughly related to pupularity * history) or about number of watchers.

Orkut, of course, ranked people by the number of "friends" they had. Or still does I suppose, but since it's a closed world, since you can't link into individual pages because of the need to log in, it's not really part of the Web.

MySpace, flickr, livejournal, they all seem to have metrics for success. In the most successful sites, you increase your "score" by doing things that also help the site grow, or help the community become more tightly knit together.

So,... what would help Advogato in this way? Of course, ideally I'd like to ask the people who left, and who are probably not reading this, why they wandered off. Or, what might make them return.

What features are most important in order to support a community?

Missing features , posted 30 Oct 2006 at 06:15 UTC by StevenRainwater » (Master)

I'm very interested in this subject too. In a few cases, as I've looked at old user profiles, I've noticed that users who left Advogato made a final diary post mentioning where they're moving their blog and why. Several mentioned specific features Advogato lacked or complaints about something on the site. It would be nice to find all of these and tally up the results to see what the most commonly mentioned problems were.

I'd like to start a list of specific improvements to Advogato that would make the site more useful to the current users and might tempt some of the ex-Advogato users to return.

I've already collected many of feature requests for mod_virgule and compiled them into a mod_virgule ToDo list. But this list is not really specific to Advogato. Many of the features were requested by other users of the software.

Thoughts, posted 30 Oct 2006 at 21:34 UTC by cdfrey » (Journeyer)

In a way, I actually like the fact that there are no comments to diary posts. This means that people posting responses to other diary entries need to link to what they are replying to. Also, everything a person writes is located in one spot. There is no way for someone interested in me to find out what I wrote in all my Blogger responses, but they can easily see it on Advogato. And by looking at my responses, they are introduced to other people that I also find interesting.

Of course, the opposite direction doesn't work. Reading an original post in my diary will not display the replies to it.

And that is a feature that I believe advogato can fix technically. Instead of writing a reply in my diary, and placing an href link to it, it would be helpful if there was a "traceback" link option, perhaps <original>cdfrey/43</original> to reply to cdfrey's 43rd diary entry. This would place a link in the response entry as usual, plus add a traceback link to the list of replies in the original.

This feature could possibly be exposed via RPC to support replies from external blogs, although that might be a problem for spam. It would work well within advogato though.

And once this feature was added, it would be neat if replies to advogato articles (like this) also showed somehow in a person's diary. Being able to find everything one person ever wrote, in one convenient location, would be very useful, in my opinion.

More flexible blogging?, posted 31 Oct 2006 at 17:40 UTC by StevenRainwater » (Master)

I went through the first 250 or so users on the user list and located quite a few ex-users who posted some sort of final good bye message. From my thoroughly unscientific sampling, it seemed to me they could roughly be divided into two groups:

  1. Advogato users who wanted to continue keeping their blog here but didn't because Advogato lacked a particular technical feature they wanted (e.g. blog comments, photos, atom feeds, etc.). Often they didn't say exactly what feature was lacking. These users mostly seem to have gone to other blogging sites like blogger.
  2. Advogato users who simply wanted to operate or host their own blog and didn't have any convenient mechanism to syndicate their blog posts to their Advogato account.

I suppose I would have fallen into the second group myself, except that I wrote some perl code that syndicates my blog to my Advogato and accounts (this was before the XML-RPC interface was added to mod_virgule).

The Planet sites, which were inspired by Advogato's recentlog, are based on a feature Advogato doesn't have: they can aggregate posts from the user's blog. Maybe adding something like that here would encourage more dormant users to come back and add their blogs to our recentlog again? I'm guessing this would at least gather back those users who are now aggregated at planet (former) advogato.

Maybe a new field on the user account page where you could specify the RSS URL of your blog and request that it be syndicated here. To prevent more blog spam, we'd have to limit the feature to trusted accounts, of course.

Some notions, posted 31 Oct 2006 at 23:59 UTC by salmoni » (Master)

I think the lack of syndication was a major reason for people leaving. Many wanted cool features like having graphics or comments so went elsewhere, and a syndication would have enabled them to stay in touch. We would need a link to the original post too in case we wanted to comment on something and the author didn't come to this site any more.

A way to edit articles would be useful too, though this could be abused. However, I've often noticed glaring mistakes in my articles which I don't know how to correct - one was a reference missing, quite trivial in terms of impact (ie, non-contentious), but important where credit is involved. To prevent abuse, perhaps an automatic post could be made detailing that changes have been made - or authors would be encouraged to post a comment to say this.

Perhaps even a continuation page for long articles, though that could be accomplished by authors submitting parts as separate articles. Does anyone feel this is not acceptable (ie, filling up the articles page with multiple "parts" of the same article)?

Steve - this really is great work and thanks very much for your efforts.

Editing posts, posted 1 Nov 2006 at 08:45 UTC by cdfrey » (Journeyer)

When I am logged in, and look at my own diary page, there are lots of little [Edit] links.

I believe when accessing these posts via XML-RPC, the date of last change is available as well, so you can detect edits if you save the timestamp with your local copy.

No Follow, posted 1 Nov 2006 at 18:12 UTC by vab » (Journeyer)

One of the reasons I used to post to advogato was to help me get my new projects out there - to get them some google juice and on people's radar. Now that no follow is in place, there's less incentive for me to post.

Re: No Follow, posted 1 Nov 2006 at 18:40 UTC by StevenRainwater » (Master)

No follow relations are only applied for untrusted accounts. If you're certified as an apprentice or higher, you can use anchor tags and get the same google juice as before.

Editing posts, posted 1 Nov 2006 at 19:31 UTC by StevenRainwater » (Master)

cdfrey, I'm assuming salmoni is talking about posting articles rather than diary/blog entries. As you noted, it's possible to edit a diary but not an article. Article editing is something we've needed for a while. Patches welcome. :-)

Patches, posted 1 Nov 2006 at 23:16 UTC by cdfrey » (Journeyer)

Speaking of patches, the CVS tree linked to from the Show Me The Code page seems to be about 2 years old.

I probably missed the correct link somewhere in the announcement, but maybe the code page could be updated as well?

Show me the code..., posted 2 Nov 2006 at 02:12 UTC by StevenRainwater » (Master)

I've updated the code page to link to my mod_virgule web page where you can download a tar gz of the current source. I've got subversion installed over here but haven't had time to play around with it and figure out how to make it go. Maybe soon.

I'm a day or two away from uploading some new code that has a lot of changes, so you might want to wait for that before doing anything major.

Looks like it may not be too difficult to get some basic blog aggregation going. DV has already done the hard part for us by adding nanohttp to libxml2, so I can grab an RSS or Atom file with two lines of code. I've already got some test code working. Just need to clean it up and do some real world testing.

extending the trust metric, posted 3 Nov 2006 at 18:50 UTC by rillian » (Master)

One thing I've been wanting for a while is the ability to cert rss urls. That breaks open the closed garden, and would let mod_virgule function as planet software without the feed moderation overhead. This is the complement of being able to import feeds as a diary (for persons and projects!). Together these two features would make it much easier for users to move on and off advogato as a hosting site while remaining part of the community

But this has turned into an advogato feature discussion, which wasn't Ankh's question. Advogato was one of the community spin offs from slashdot. When that site got big enough for the level of technical discussion to degrade, a lot of people were looking for another place to carry on conversations. Advogato's initial popularity was due to that, but these days no one has the luxury of being the only one in a particular niche. But just being more active and open would help, I think.

Cert RSS?, posted 4 Nov 2006 at 10:17 UTC by salmoni » (Master)

rillian - certifying RSS urls sounds interesting. It's a shame that RSS itself can't handle it independently. That sounds confusing so I'll try to explain:

Imagine if you were looking for interesting things to read, you could simply use a search engine* to find RSS feeds that have been highly rated by others. This is independent of any website so is like /.'s moderation but on a wide scale. We could do things like searching through RSS news sources and filtering out low rated feeds, or comparing news sources (eg, bloggers against more "authoritative" websites, or BBC vs Reuters vs Fox). Of course, the problem would be spoofing by ego-whores or spammers making their RSS feeds highly rated. Perhaps this could be an application of the trust metric?

lol! I just like the idea of the wider Internet community rating the authority and "worthiness" of different feeds. Note - I'm aware that aggregation sites do this, but these are awkward for most people to find. Just imagine that you have an RSS reader that shows you the feed and has a rating at the bottom. You rate it (along with everyone and anyone else who wants to), and this rating is visible to anyone else who finds the feed. Is this clear?

Another problem - where is the information stored? Having a central location seems to limit the utility, and would inhibit take-up (unless it was from an established company like Google); but the only other way is to mess with RSS itself which is unlikely to happen.

* Does Google do RSS only searches? if not, why not Google? Having RSS only searches could be useful even without ratings and would probably be easier to use than searching through piles of aggregation sites.

Gaming, posted 5 Nov 2006 at 22:52 UTC by Zaitcev » (Master)

I do recognize and do not care for site-imposed "metric for success". I was not aware that LiveJournal had one. I may belong to the minority in my indifference, but I am quite sure that trying a reward scheme to enliven the recentlog is not going to produce anything as interesting as the old recentlog. The whole point of it was a snapshot of a certain active community. If community itself evolves away, there's nothing that can be done by the site software.

I considered my membership at Advogato as a give-and-take between Raph and myself (highly asymmetric, of course): I wrote entries and read recentlog, whereas Raph received material for his research. But the score-driven gaming was not very strong at the site, and I consider it a good thing.

By the way, many people tried to award points for things like filing bugs and posting patches. For example, in Linux kernel there was an effort to assign points to authors of netfilter patches. As far as I can tell, all such efforts fizzled out.

So, the artifical community building seems rather masturbatory to me. If Steven and Rillian were to stick together and deliver some significant application of trust metric, that would be interesting for me to be a willing guinea pig. Otherwise, purpose-built blogging sites do a better job.

Guinea pigs, posted 6 Nov 2006 at 22:16 UTC by StevenRainwater » (Master)

Speaking of guinea pigs, Zaitcev, I see that you're hosting your blog at Live Journal these days. How about being a guinea pig for the new blog syndication feature? It's not finished yet but the aggregator back-end is working and I could use a couple of test subjects who have Advogato accounts but don't keep their blogs here. All you'd need to do is go to your Advogato account profile page, check the "Syndicate your blog from another site" box and add the URL of an ATOM 1.0 or RSS 2.0 feed. Once the coding and testing is complete (maybe one more week), your Live Journal blog posts would appear here on Advogato and show up in the recentlog.

rillian, how do you envision your RSS certification scheme working here on Advogato? As currently implemented (well, partially implemented) each trusted Advogato user can choose to populate their diary from an RSS/Atom source rather than posting directly to Advogato. This seemed like the safest way to start since it uses the existing diary/recentlog/diary-ranking system., posted 6 Nov 2006 at 23:00 UTC by lkcl » (Master)

this goes back some time - to 2000 / 2001. integration of into advogato would, i believe, be a significant step forward.

conceptually, certification of rss feeds doesn't exactly work - but the use of openid does.

i don't believe that certification of rss feeds has any actual meaning. after all - a trusted advogato user is going to be very very stupid to actually add someone ELSE's rss feeds to their own account.

which is where, if necessary, openid would come in: if there is a mechanism for validating that a particular RSS feed comes from a certain user, then a flag could be raised, on the site, saying 'advogato has checked [the digital signature of] this rss feed and it belongs to the same openid user - and if you don't believe us here's how you can check it for yourself'.

Cert RSS, posted 7 Nov 2006 at 01:22 UTC by rillian » (Master)

StevenRainwater, I hadn't thought about it in too much detail. I imagined you'd have a web page where you could paste an rss feed url. This would create a special account instance which you and others could then certify. The rss feed would then show up in the recent log, subject to the normal filters, and the account page would link back to the blog or website, perhaps with that url entered separately.

People could later "claim" the advogato account through some kind of "make this token appear in the feed" mechanism, and then be able to get a password and change the account nick.

It's an extension to what you're working on with importing diary entries from rss, but without the assumption that the owner of the feed will create the account first. Instead, if you know someone, you can just cert their blog feed and they become at least of leaf in the trust metric. A person could do this manually by creating an account for someone and hooking it up to their feed, but that's not very friendly if the target individual later wants to assume control of the account. Thus, advogato should support that directly.

salmoni and Zaitcev, I've talked about distributed trust metric calculations some with Raph. It's hard. He did some work on a P2P system that did this implicitly (search for stamp trading network) but nothing much came of it. The currency idea is an interesting spin. Spam filtering by Capitalism! I think it's more realistic to just harvest links and use them to do a local trust evaluation until we understand things better.

As far as rss feeds certing each other, they do through their links, although this isn't quite the same as the assertion advogato uses. But you could harvest new feeds by looking in the pages your known feeds link to. Scraping blog rolls would get you certs closer what advogato uses. "People worth reading" and not just "article I wanted to talk about".

Providing "single sign on" was another thing we talked about. That's entirely my fault; I said I'd implement the scheme developed through the recent log and never did. Nowadays, as lkcl pointed out, there's openid, so some kind of integration there would make more sense. I hadn't mentioned it because I saw it was on the todo list

trust metric implementations, posted 7 Nov 2006 at 17:04 UTC by lkcl » (Master)

with respect, i disagree that distributed trust metric calculations would be difficult.

with the current use of the ford-fulkersson 'maximum flow' algorithm, which is a depth-first algorithm, hell yes, it would be difficult.

however there are plenty of other max-flow algorithms, some of which are breadth-first, and it is breadth-first algorithms that would, along with cacheing of certs from other sites, prove much more amenable to distributing.

the other important thing to have when going distributed is digital signatures on things, otherwise it's no good doing _anything_ like this except as a 'nice experiment'. we implicitly trust that nobody hacks in to and changes the database. however, a remote site must also be trusted: by the time you get to 100 remote sites there's no way you can rely on every site to be trusted (not hacked) so instead you really need to use digital signing.

that's what 'keynote' was all about (look it up with google).


sorry, rillian: i assumed that you meant 'actually certify the rss feed'. what i believe you _actually_ meant was 'create an advogato account with an identity and then add an rss feed to that which over-rides the advogato diary' is that correct?

and then people can certify the advogato-user-which-happens-to-have-an-rss-feed-as-a-diary-instead-of-an-advogato-diary


Why I haven't posted much, posted 7 Nov 2006 at 18:42 UTC by slamb » (Journeyer)

I don't think providing a score would help matters much. My two reasons:

First, a technical one. The recentlog is no good for carrying on conversations. I try sometimes, but people miss my replies, and since the recentlog only shows one post per person, posting again means giving up on the old post being read. So if I've replied to someone recently, I have a disincentive to post.

Second, a personal one. I've been busy, and generally not with free software. I haven't stopped entirely - and I guess there are a few things I could post about now - and I will surely start writing more again. But advogato's had a shrinking crowd for quite some time, and everyone goes in streaks. I think the only way to consistently get lots of good posts again is to get a larger crowd again.

I could post about what I had for breakfast, or proprietary software I write at work, but hopefully if I did that everyone would mark my diary as boring and you still wouldn't see it. One of the things I like about advogato is that it's focused. I only read the entries that talk about technical things. I like to think other parts of my life are interesting, but this really isn't the place for them. Here free software is signal, all else is noise.

More distributed posting, posted 7 Nov 2006 at 21:10 UTC by rillian » (Master)

slamb, the one-post-per-recentlog has been fixed. Yay!

lkcl, something like that. What I said is what I wanted to do conceptually, from the point of view of user experience. Instead of certifying the identity of behind an advogato account, I wanted to certify the identity behind an rss feed. But within advogato's framework that has to be implemented by a class of account, as we both said.

Relying on digital signatures still amounts to trusting that "nobody hacks in...and changes the database." But there are other ways to enforce that sites can only make certs from nodes they control, say through the dns system or probabalisticly unique ids, and the trust metric should still let you distinguish between trustworthy and untrustworthy remote sites and their nodes.

At least, I think that was the point of the distributed discussions. Parallel algorithms where you don't entirely trust the nodes doing the computation is the hard part of the problem.

Attachment hosting, posted 7 Nov 2006 at 21:16 UTC by rillian » (Master)

On the thread of making advogato more attractive, we now allow <img/> tags, and I think so far they've been used to great effect. However, there's a drawback for someone trying to use advogato as their primary blog, in that the images must be hosted off site, even though they're supposedly part of your blog. Not robust if you're posting other people's pictures, and you still need another site if you're posting your own.

I don't know what kind of resources our new host can provide, but I'd suggest that trusted accounts being able to attach and refer to files (images, media, source code!) in blog and article entries would be a nice feature.

Syndication..., posted 8 Nov 2006 at 23:28 UTC by StevenRainwater » (Master)

We're getting closer on the blog syndication. I got a little carried away yesterday and decided to add support a few more feed formats than I'd originally intended. Zaitcev's feed uses Atom (thanks for volunteering as a test subject, by the way!), so that's first on the list. A few other people have set up a feeds too, so I have sample data for the common RSS formats (v0.91 and v2.0). And one feed was in the somewhat more complicated RDF Site Summary format (aka RSS v1.0). Looks like all of them will work out okay so far. After parsing all these formats, I can see why Atom is becoming so popular.

lkcl, OpenID is definitely on the ToDo list. The big question is would it make more sense just to add support for logging into Advogato using OpenID authentication, or to provide OpenID authentication to other sites using Advogato user data. Would there be some value an OpenID provider with trust metrics? If a login succeeds, you've not only authenticated the login but also that the user is a trusted member of the Free Software community. Hmmm...

Feed for recentlog?, posted 10 Nov 2006 at 03:43 UTC by Alphax » (Apprentice)

Either I'm thick and can't find it, or there isn't one. If there is, where is it, and if there isn't, can we get one?

Feeds, posted 10 Nov 2006 at 16:32 UTC by StevenRainwater » (Master)

Alphax, I've added a recentlog feed to the ToDo list. I've also added a couple of other feed-related tasks to the list. I think we need to move all our existing RSS feeds to RSS v2.0. I'd also like to look at making Atom feeds available as well.

distributed trust, posted 10 Nov 2006 at 17:23 UTC by lkcl » (Master)

ok. i believe that you believe that the problem is harder than it is. you can always run the trust-calculation program either yourself, or you pick a specific server that you trust to perform the calculations for you.

then, that server (or you run the program yourself) performs a breadth-first scan-and-download of digitally-signed trust-metric-certificates.

if the trust-metric-certificates are digitally-signed (via a mechanism such as GPG which can be verified or by i presume openid can do digital signing?) then you have a chain (rather a large and hierarchical chain) to reassure you that the resultant trust-metric-calculation comes from the right sources.

the point of using keynote is that it gives the next clue: the resultant trust-metric-calculation _itself_ is then digitally-signed (!) and then that calculation itself you can double-check and verify that it was performed in good faith.

it really _isn't_ rocket science.

i'm assuming of course that slurping all of people's certifications onto one server, breadth-first, is a practical option...

... however - if you are talking about millions or billions of certifications, then yes you have a serious problem: that _does_ need some rocket science and a totally different approach.

if you're talking about actually distributing the trust metric calculations, then yeh, that's a big hairy deal.

Re: distributed trust, posted 10 Nov 2006 at 17:44 UTC by rillian » (Master)

lkcl, that is an excellent summary. Thanks!

message to people creating a new account, posted 11 Nov 2006 at 07:47 UTC by lkcl » (Master)

steven, hi,

i think it's important to advise people, now, on the create new account page, that they fill in some background information about themselves, and say that it is important that they do this, so that people can establish who they are and can Certify them.

a genuine user will do this.

disingenuous users will ignore it and consequently will be easy-to-spot as spammers.

sneaky bastards spammers who start stealing genuine free software users' identities is the next serious problem up the chain to think ahead to that could hit advogato: that can easily be dealt with using openid or gpg.

Status update - 2006/11/25, posted 26 Nov 2006 at 04:03 UTC by Zaitcev » (Master)

The syndication seems working swimmingly, now that it has the back pointer. One thing I noticed though... Old entries are not being updated if changed upstream.

updated entries, posted 26 Nov 2006 at 21:13 UTC by StevenRainwater » (Master)

Zaitcev: I've looked at the syndication formats and Atom has a handy "updated" tag for just that reason. It shouldn't be too hard to add support for that one. Handling updates on RSS or RDF feeds looks a bit more involved. I don't see any obvious way to uniquely identify a post or to tell whether or not it has changed.

lkcl: Yes, I think you're right about needing an update to the account page text. Several other Advogato pages could use updates as well. I may have time to take a shot at it this week sometime.

Looks like we may need to add a "delete blog entry" feature too. I cleaned up gotcha's little recentlog accident manually but it would probably help if users could delete their own blog entries when needed. I've noticed other users who, for reasons I don't fully grok, replace their older blog posts with empty entries to effectively delete them anyway. So I think there are multiple reasons for adding a delete feature.

less-recent log?, posted 2 Jan 2007 at 04:20 UTC by Alphax » (Apprentice)

It's a pity the recent log runs out so quickly... the ability to hit "older" (as we can with articles) would be nice.

Re: less-recent log, posted 3 Jan 2007 at 00:18 UTC by StevenRainwater » (Master)

Alphax: Adding an "older posts" link at the end of the recent log is already on the ToDo list. Now that holidays are over, I should be able to find some spare time soon to start coding on mod_virgule again.

This is a problem we should be happy to have, anyway, since the article at the top of this thread was complaining that the recentlog had gotten too quiet. I'll take this as a sign we're begining to turn things around.

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