MySite - Social Architecture & Planning

Notes from Venky Veeraraghavan's presentation, forgive the typos!

• Session objective: understand technical architecture and planning required to deploy
  • MySite personal portal
  • Social feedback
  • People and expertise search
• @SPSocial is the Twitter ID to follow
• Demo:
  • MySite has 3 out of box pages ... expect people to rename pages to make them their own
  • Shows stream of comments on Note Board and activities re: tracking people and tags
  • Activity feed is very extensible stream ... can track items coming form within SharePoint but you can build connectors to include other sources
  • Will talk about mining Outlook later
  • Shows Silverlight org chart browser
• Major components
  • Pages: My Network (was in 2003, removed in 2007 now back?), My Profile (was in 2003), My Content (was in 2007) (own personal site collection, you are the administrator of your own site), My <foo>
  • Site Collections: MySIte Host (centrally managed), My Site
  • Web App: My Web Application
  • Service Application Proxies: EMM (meta data). UPA, Search
• Activity Feed
  • New for SP2010, web part, Atom 2.0 feed
  • Consolidated: activities from everything I track (inbound)
  • Published: my activities (outbound)
  • Should be able to export atom feed for other feed readers
  • Need to consider performance implications of feeds
  • Architecture: two databases: social data (note feedback, social bookmarks, social tagging) profile database (profile properties, membership colleagues etc). "Gatherer" job scoops up data and sends to feed table to scale ... simple select all items I track ... then API to inject into feed store ... you can write gatherers that collect data from sources you think are interesting and then stick into feed - set to 1 hour timer
  • Note: permission models! What are the ACLs around the feed elements so that right people see only permissions items in consolidated feed. Also true for "My Activity" stream - how to wrapper what is sent outbound.
  • Extensibility requires some key points to know: Activity Application (say CRM), Activity Type (say New Meeting) and then an Activity Template "<person> has scheduled a meeting with <customer> on <details> for user experience.
  • Activity Event: "Mike has scheduled a meeting with Company ABC on DD, MM, YY" is displayed
  • Considering whether you can "friend" a gatherer if I heard it right...
  • Consumption API is an Atom 2.0 feed
• Feed Readers
  • Goal of 2000 RPS current at 500-700 with activity, 2100 without activity RPS = requests per second
• Profile Pictures
  • Central picture library
  • Resided three ways (for use in SP 32x32, use in AD and client apps 48x48 and for profile page 96x96)
  • Controls needed for photos due to legal and risk concerns (e.g., harassment) 
  • Picture picker, supports policies, can replace with other picker
  • Sync up to AD, supports write-back if configured
  • Outlook and OCS use Pictures in AD
  • Note: really, really, need to plan for privacy
  • Top issue: Picture usage: consent, opt-in, corporate policy
  • Top Issue: Activity feeds: who follows me and possible need for two-way consent
• Bookmarklet for browsers, can see who else in company has tagged page, discussions around URL, using Firefox in demo,
• May also want to tag customers, may have LOB application for instance, can tag customers or other artifacts - you will want your info to be URI accessible thus leveraging tags
• Note: Wonder if there will be a bookmarket for Office client tools not just browser.
• Social Feedback: Social tags, notes and ratings
  • Helps categorize, annotate, promote and help retrieval of relevant links
  • Applies to any URL inside or outside of SP with bookmarklet
  • Independent of write-permissions
  • Primary way or promoting documents and web pages to news feed
  • Basic level 3 tuple: Person, URL, Feedback
  • Note: still need permission controls around sources being tagged/bookmarked as well as tags/bookmarks - not sure enough security was discussed and therefore not sure if it is in there
  • Note: biggest barrier: not enough intranets have rich native web content
• Social Feedback table has 3 tables with foreign keys into Profile and EMM (Enterprise Metadata)
• EMM supports folksonomy and taxonomy terms so people can select the meanings they want
• Now we get to security ... need to prevent tags as giving away existence of document people don't have access to
• Security trimming ... not explained deeply, pointed to other sessions
• Related services: search service index tags and security trimming
• Scale: Cab be very large datasets ... enterprise metadata generates tag at Internet scale
• SP2010 will estimate 600 million rows at RTM - use Excel model ... social feedback database
• Need a privacy policy for tags/bookmarks ... transparent sharing, what happens when an employee leaves, security, activity feed repercussions, etc
• Note: security teams will look at this to see if it introduces additional surface area for attack as well as data loss/leakage etc
• Need to plan for adoption - employee advisory committee prior to deployment, seed the network with tags, connect with HR, legal, execs, etc
• Workflow for handling concerns and escalations
• Mining Outlook for expertise is consent driven ... nothing every published without user concent
• Knowledge Mining: Tags are in Outlook 2010 and SP2010 (knowledge network stuff), client does analysis, get consent on client and server, indexing occurs and they can query
• Colleagues (social graph) is indexed as well as distribution list and site membership (see Social Search session)
• First question people ask is "how did it know" (the system) so user education about company policy is important, "suggestions" form system only visible to owner, only indexing profile properties marked for "everyone"