I've proposed an alternative RESTful batching format to the OpenSocial spec list. It's conceptually the same as the multipart proposal, but hopefully much easier for web servers and clients to deal with (JSON only for now).
...up in Sebastopol. May actually camp this time. There are so many Google people going, I looked into whether we could charter a bus, just like summer camp. No dice; maybe next year people will see a bunch of Googlers singing songs in a bus.
I went to a talk recently by Carol about how children's attitudes towards performance and learning greatly affects their performance, with some good statistical data on studies she and others have done. The take-away, for children: It's more more effective to believe that you can and should grow your abilities vs. believing that abilities are inherent. Don't tell your childen they're smart (fixed mindset), tell them it's great that they're trying hard and can improve (growth mindset).
The problem with the fixed mindset is that it leads to fear - if you can't do something right, then clearly you can't do it, and you'd better hide that fact as quickly as possible, mostly by not trying. The growth mindset, on the other hand, assumes that you can always improve. The latter, unsurprising…
See Assaf's analysis. Is it totally automated? Or generated by humans? (And does it matter in choosing defensive strategies?) Text with the only suspicious item being a link to a suspicious site is more difficult to deal with -- the referent may be fine when you check it, but change to an SEO spam site two days later, for example. And just checking web sites is expensive too.