Dominic Sayers

Social Data Flow minor update

I'll probably write some more about Social Data Flow shortly, but for now I just wanted to talk about extending the canonical data flow with bespoke updates flowing through my own site. In a perfect world, my syndicator (which happens to be ping.fm) would take care of all my syndication needs. It would take my micro-blogs and status updates and send them to all known social sites where I have an account. Unfortunately while ping.fm appears to have the best coverage, it's not complete. Posterous, for example, is a significant omission (any site where Stephen Fry is active has to be considered a significant site, or anyway it soon will be). As it happens, it's very easy for me to update Posterous with my ping.fm outpourings since I've already implemented a ping.fm Custom URL for my own site - it updates the dynamic sections in the left-hand column. All I had to do was arrange for the PHP page that receives the ping.fm updates to then mail the same update to Posterous. Job done. Not sure why ping.fm haven't done it themselves, except that pressure of other feature request probably prevents them doing everything they want to straight away. More to come on my struggle to get an acceptable flow of location data in place.

Social data flow in action

OK, I've seen a few things online recently that amused me and I'd like to share them. I don't want to add them to delicious.com because that's what I use for links I need to find again someday. These are things I've seen and enjoyed but they are gone. For me they are yesterday's newspaper. But you may not have seen them and I think you should. So I've decided to add them to Tumblr. Easy. Tumblr is a great tool for doing this, it makes it easy to add videos, links, pictures, anything. I am very happy with this choice. But this has now repurposed Tumblr in my social data flow from an output device to a content provider. Before, it just used to receive my status updates and micro blogs. Now it is the place where I link to things that have amused me. Here's how I needed to reconfigure my social data flow: 1. Add Tumblr to my aggregators. Now my Tumblr posts will appear in my All Activity feed. 2. Remove Tumblr from my syndication list. Now Tumblr will not receive my status updates or micro blogs. This is important because otherwise these updates would appear twice in my All Activity feed. 3. Er... 4. That's it. Simple.

Revised social data flow

In an attempt to remove the latency problem from my social data flow, I've rearchitected it a bit. Unfortunately the reworked flow is not quite as elegant as the original but then it had more sophisticated design goals so that's not terribly surprising. Here's the revised flow with notes below:

(Click on the image for a bigger version)

Notes:

1. Now I can allow ping.fm to update Twitter and identi.ca immediately so I have removed the latency problem from the previous configuration. Once again I can use Twitter in a conversational way instead of it being a mere receptacle for RSS updates every half hour.
2. You'll note also that Twitterfeed can now update identi.ca and other Laconica-based micro-blogging sites. Thanks, Mario. This means I can promote identi.ca to be a peer with Twitter and receive all my updates including real-time micro-blogs (tweets).
3. The three subscription points are the same: Writing, Status Updates and All Activity. The URIs for those resources have not changed.
4. I have distinguished between blog posts such as this and the notification of those posts. To subscribe to the full post stream use the Writing feed. If you subscribe to the All Activity feed you will be notified about posts but only see the title not the whole article. Unfortunately if you subscribe to both you will see both the blog post and a separate notification of its existence - I can't do anything about this yet.

So there's an emerging gap between those select sites that I use to do stuff and those sites that merely receive notification of that activity. My strategy will be to keep to a minimum the sites I use for doing things, and maximise the number of sites that are notified about it. This way I can keep all my pictures, links, writing etc. in one place but all my disparate multiple friend communities will get equal notification of new activity in one of those upstream sites. I think the new slightly less elegant social data flow is functionally better and I'll try to stick with it for a while to see how it performs in practice.

Initial thoughts on my new social data flow

Firstly, apologies to anybody who couldn't see the social data flow diagram in my earlier post - I've now hosted it on my own server so it shouldn't disappear quite so often.

(Click on the image for a bigger version)

The new data flow, using ping.fm to syndicate my status updates and micro-blogs and FriendFeed to aggregate my other activity and consolidate everything into a coherent set of feeds, is working quite well. I would like to be able to have more control over the format of what both ping.fm and FriendFeed do for each target system, but the defaults are adequate for now. Here are the issues I'm experiencing: 1. Using Twitterfeed to take my FriendFeed RSS and pipe it into Twitter introduces a degree of latency that's a bit inappropriate for Twitter. Twitter should be near real-time and conversational (whilst not being an instant messaging system), but Twitterfeed only reads my FriendFeed every half an hour. If I get ping.fm to update Twitter directly then I'm going to get some updates twice, which is exactly what I'm trying to avoid. 2. Some of the connected systems (identi.ca for example) are Twitter-like, and I'd like to treat them the same way as Twitter: as an output device for all my activity. But Twitterfeed is Twitter-specific and I haven't found an equivalent for identi.ca etc. yet. In any case I don't want a different Social ETL tool for each output system; I want Twitterfeed (or something) to do this for a number of target systems. 3. There's no class solution for location-based systems yet. Obviously I'd like to be able to update my location and travel plans in one place and have it propagate to TripIt, Dopplr, BrightKite etc. I think FireEagle may be trying to do this but it's short of two things at the moment: co-operation from the other systems and an invitation for me. If anybody has access to FireEagle I would appreciate a way in. 4. Now I need to do the other thing - deduplicate my profligate friends' updates. When Myrto uploads a picture to Flickr, for instance, I get notified about it four times. I can rationalise this a bit but I can't ignore her FriendFeed or Twitter because I would miss some of her updates, so I'm condemned to hearing about her new pictures multiple times unless (a) she adopts a version of my social data flow or (b) I find a solution to deduplicating the same update coming through multiple channels. This duplicate filter is necessary until everybody adopts my architecture, i.e. for ever. Does anybody know of a solution out there? This all seems quite difficult to manage and well beyond anybody who is either busy or technically challenged. There's no real way of developing a packaged solution until all the social systems adopt a single sign-on technology like OpenID, and many of them have sound technical reasons for not doing so (OpenIDs can be created by anybody anywhere so it's a bit like expecting system owners to trust a digital certificate that has no trusted root certification authority). Anyway, that's the state of play at the moment. More to come I'm sure.

Item status

Richard Rutter has come up with a better phrase to describe the Read/Not Read status of a feed item. He just calls it "item status". That will do for me from now on. He chose Shrook as his aggregator because it supported synchronisation of item status between environments back in 2004. Unfortunately it's just a Mac client. However, it's good to have company in recognising this is a valuable feature.

Social OPML

Remember how you used to save your bookmarks in the browser (Favorites if you use Internet Explorer)? Then when you discovered del.icio.us you realised that you no longer had to maintain your bookmarks in two different browsers on three separate PCs. And you could use the fact that everybody else kept their bookmarks in the same place to make connections between your bookmarks and other people's You don't go back to using local bookmarks once you have seen the world of social bookmarking. Browsers have been around a lot longer than feed aggregators. Maybe that's why we are still happy to keep our feed collections locally. Maybe we simply haven't discovered social feed collections yet. But this would be the answer to something that is annoying me a lot more now than it did a few months ago. As more and more of the information I consume is available in a syndicated format, I add more and more channels to my aggregator. Some important messages now come through my aggregator. I am starting to depend on it for critical notifications. Wherever I go, my aggregator is open on my desktop and it is as necessary to me as my email client and my chat client. So the fact that I have to re-read the same posts that I have just checked when I move to a different PC is starting to bug me. The fact that I have to add a new channel manually to each aggregator on each PC is starting to bug me. I want to keep my feed collection centrally, outside my aggregator. And the Read/Not Read status of the posts too. del.icio.us for feeds. That's what I need. You read it here first. Stop and think about it for a minute before I start talking about implementation problems. 1. There is, of course, a chicken and egg issue. There's no point storing my feed collection centrally unless my aggregator can use the central store. The central store won't get any kind of critical mass until most common aggregators are using it. None does at the moment. 2. The second point is that once you have decided to store your feed collection centrally and also the Read/Not Read status for each post then you are very close to deciding to store the posts themselves centrally. I'm not sure about this although it seems to be a logical conclusion. 3. I don't understand the business model of del.icio.us sufficiently to understand how you could finance a feed.licio.us equivalent. I hope these are mere details :-)  If we build it they will come, as somebody once said. p.s. This post arose from a conversation with Adam Tennet

Corporate feed aggregator revisited

I should clarify something I said in a previous post about the ideal corporate aggregator. I said it could not be browser-based and when I said that I had in mind a very specific architecture that would not work in a corporate environment. Thinking about this some more there are of course ways of implementing a personal aggregator with a browser interface. My objection was to feed caches that simply subscribe on behalf of their users to a set of feeds then serve them using HTML. This cannot work in a corporate environment because the server is either subscribing only to public (non-secured) feeds or supplying a general set of credentials on behalf of the entire user population. You could implement a corporate aggregator with a browser interface if you designed it so that it was in effect a virtual personal aggregator application sitting on a server. Each user would have his or her own credentials held on the server, and for secured feeds they would need their own personal storage of that content. Public content could be shared among all the users and this might cut down on the overall storage requirement across the corporation. How many times do you need to store the same Boing Boing posts? So, I will reiterate my seven requirements for an ideal corporate aggregator: 1. It must be able to supply to a secured feed the personal credentials of an individual user. 2. It must support all the authentication mechanisms used in the corporate environment: digital certificate, Kerberos, NTLM. Any more? 3. It must be able to be managed centrally by the support department. This means managing a user's portfolio of feeds for him. There are three reasons for this: (a) there are users who will not want to manage their own feeds at the outset as the technology is new to them, (b) there will likely be a standard set of feeds that all users must consume (corporate bulletins and the like) and (c) when the aggregator is upgraded or replaced we will need to preserve each user's feed collection during the process. These actions need to be scriptable in some way. 4. The Read/Not Read status of any post in any channel must follow the user around. Corporate users change locations frequently, they work from home, they make VPN connections from internet cafes. Whatever the situation they need to see the same feed collection and the same unread posts. 5. The user interface must be sufficiently rich to make consuming a large number of feeds efficient. This is another challenge to browser-based aggregators which currently do not support single-key reading. However carefully you select your channels there will always be a certain amount of noise in amongst the signal. You need to be able to move rapidly through irrelevant posts to give you more time to read the relevant ones. 6. On the same subject of signal-to-noise ratio, there are channels that frequently have useful content so you need to subscribe to them but it is mixed in with the personal bugbears of the author. Some feed publishers will provide a feed of certain categories of post so you can filter out the noise this way, but this means you miss any interesting new categories the author introduces. What I need is a way of subscribing to all posts except those which mention Unicorns for instance. The aggregator must support filtering. 7. We need to get away from the arrangement of feeds into a folder hierarchy. I want to tag my feeds with arbitrary categories. I want to be able to Mark All As Read in any given category if I am pushed for time. Der. I can't believe nobody has done this yet. Tags not hierarchy! Somebody tell me they are working on this, please.

Blog post filtering

I agree with Cory Doctorow for once. Feed Rinse is great. Feed filtering should be built into your aggregator. My €0.02, which is why I'm posting rather than just adding the link to del.icio.us, is from my standpoint as an evangelist for a corporate aggregator. If you're consuming internal feeds (i.e. feeds from your intranet) then you can't use a separate, external third-party filter. If you're consuming secured feeds (feeds that start with with "https://") then you can't add an external filter into the data flow without moving all your authentication credentials into the filter service.  It has to be built into the aggregator.

Features of a corporate feed aggregator

Here's what I think is essential for a corporate RSS reader: 1. It must support individual authentication in order to access secured feeds. Preferably using a digital certificate to identify the user. This rules out all browser-based aggregators 2. Most users will want to get started straight away and will not want to spend a long time managing their list of feeds. We need a way of managing centrally a user's OPML file and adding a feed for them if they request it. 3. We need the Read/Not Read status of the content to follow the user around, whether he is using his regular desktop machine, working from home on his laptop, or in an internet cafe using a server-based computing environment or a VPN. 4. The consumption of new content needs to be efficient. My aggregator supports Space Bar single-key reading. I just hammer away at the Space Bar until I find a post I'm interested in. This enables me to consume a large number of feeds in a short amount of time. We will be accused of creating an UNproductivity tool unless we can demonstrate this! These things are pretty important for this technology to get accepted in the corporate environment. For my own benefits I would also like to see a feed aggregator that has the following feature: 5. Feeds are organised by tagging them, not by creating a hierarchical folder structure. Individual posts could also be tagged.

Books I should have read

For no other reason than self-flagellation I'm going to start a list of books I should have read. Hopefully this will provoke me to get on with it. Worse, I have copies of some of these and have still not got round to reading them. I will also note the must-read books I started then gave up on.

Some of these may date quickly. If I'd been compiling this list in the 70s I would probably feel I had to read The One-Minute Manager and stuff like that. Hey, there's another list! Books that are no longer required reading! The Mythical Man-Month, Men Are From Mars Women Are From Venus, etc. etc. If I procrastinate long enough many of the books on my list will drop off due to no longer being trendy (No Logo springs to mind as a candidate).

How can I manage this list? I would like other people to prioritise them for me since I am unlikely to be able to read them all anytime soon. Which do you recommend?

Where is the last.fm for books?

1. The Tipping Point: How Little Things Can Make a Big Difference, Malcolm Gladwell
2. Blink: The Power of Thinking Without Thinking, Malcolm Gladwell
3. The Wisdom of Crowds, James Surowiecki

Many, many more to come.