specifically regarding the RFP: I would narrow the "wish list" section down to specific items that are to be delivered. I'd also categorize what needs to be "right from the start" and what needs to be "good enough" - Tim Berners-Lee in designing the web knew that the URL syntax had to be right from the start, because it would be forever embedded in links. HTML and HTTP, however, only had to be "good enough" to get the web started, after which they could evolve with newer versions. Had he tried to overengineer HTML and HTTP in the beginning, he would have run the risk for creating a backwards-compatibility nightmare.
re: the architecture in general: My (admitedly fuzzy) thoughts on how this would work would focus on specific actions... the things being done. These are the "packets" that flow end-to-end through the network, analogous to the information packets that are routed through Internet Routers. If a node on the P2P goes down or becomes untrustworthy, then actions are "routed around the damage" in the same way that internet packets are "routed around" the Internet. Similarly, if a node proves to be trustworthy and effective, then it attracts more actions. So, the network becomes self-organizing and self-propagating: Effective actions by trustworthy nodes proliferate; ineffective actions by untrustworthy nodes wither away.
The internet was also based on a "smart edges" approach... the network itself just dumbly routes bits, not caring whether the content is an email, a ring tone, music, data file, or video. It is up to the end points of the network to figure out what the data means. Contrast this with the "Smart Center" model of the traditonal "Ma Bell" network: users had "dumb" peripherals (i.e. a phone), and the network had the "smarts" for making the calls. The Ma Bell network connection quality was only as good as its weakest link, while the Internet assumed unreliable links and dynamically routed around them.
As we have seen with the internet, this open approach creates immense opportunities for innovation, pushing power out to the "smart edges." This, of course, is very threatening to the "smart center" industry, who want to sell you ring tones for one price, SMS for another, and photos for another.
The folks who are thriving in the "smart center" network model are the last ones to support "smart edges". You can read a bit about this from David Isenberg's work "The Rise of the Stupid Network" (which got him fired from AT&T).
http://isen.com/stupid.html
I would contend that we have a "smart center" model of philanthropy and development today. The intervenors/donors/NGOs/agencies have the intelligence, and the poor people being helped are the "dumb edges" who are supposed to do as directed. (e.g. wear the right T-shirts when donors from oversees show up. Imagine the catastrophic consequences of kids wearing T-shirts from donor X when donor Y visits).
Dependency is "baked in" to the smart center model.
There certainly are interventions where smart center approaches are appropriate. Eradicating polio, smallpox, or other public health measures DO involve superior intelligence at the center... but these are not the be-all and end-all of activities, and certainly shouldn't be used as sole role model for uplift.
Maybe I haven't looked deeply enough, but I don't see this end-to-end activity model in the current SA architecture.. It looks to me that it is simply a fundraising front end to existing fundraising sites. What if there is a negative result somewhere... who is going to tell the donor that their donation didn't work right?
Let's look at the Omidyar $100 million donation to Tufts for microfinance. If all goes well, it will trigger a thriving market in microfinance, and a portion of the profits will flow back to the university trustees. Now, let's look at a hypothetical situation in which the Omidyar's good intentions didn't actually have good effect - perhaps their donation actually drained capital and interest from the poor they were trying to help.
If this hypothetical situation arose, who would tell them about it? Would the trustees tell the Omidyars that their donation had triggered negative consequences? Would the professors, microfinance intermediaries, NGOs, or field workers? If someone dared to bring up negative results, would they be treated seriously, or would they be considered a rabble-rousing conspirator?
Such are the dynamics of the "smart center" network. Everyone is tied to a dependency tree from the funding source.
Reversing these dynamics is a challenge. Rather than assuming a perfect connection managed by the smart center, we need to presume that a certain percentage of actions will not work, will have perverse results, or will be handled by untrustworthy nodes. These "negative" results or expectation failures should be used for learning and conditioning the network, and be an active part of the network dynamics.
The design thus becomes a balancing act between positive and negative results.. and minimizing or constraining the costs/risks of the negative results. One way (I think, but haven't fully convinced myself) is to shrink the granularity of the interaction... lots of little activities. in NetGeekSpeak: a massively scalable fine-grained network. This would create a "long tail" distribution of lots of little interactions, (a power-law distribution -see also Barabasi's Scale-free network works
http://en.wikipedia.org/wiki/Scale-free_network ).
The key to making all this happen is to figure out a way of efficiently shrinking the packet of interaction while increasing the feedback derivable from the interaction and the chain of trust supporting the action.
gotta run now... more to come later.. feel free to blog these comments, if you like...
tom