To respond to your larger question, the RPF has been extremely adamant in their $25 price point for the base model and that's not about to change
Well, it wouldn't be the first time that the foundation's plans change
In theory, I would be fine with Model A / 256M, but I'm concerned about how long such a model would last: 3 months after release? 6 months? 9 months? For how long will it be possible to source 256M RAM chips in large quantities and at significantly lower cost than 512M? I honestly don't know.
To make it clearer where I'm coming from, this is what I wrote when I read about the first memory upgrade (from 128M to 256M - that was March):
http://www.raspberrypi.org/phpBB3/viewt ... 319#p46319
Very good news.
I am a bit concerned about the availability of Model A, though. I hope the foundation got some written reassurance in that regard. I can see a per-profit organization limiting the supply of Model A to incentivate sales of Model B (much higher margins, I would think).
At the time, I thought: Great! the minority of hobbyists buying Model B will subsidize the majority of poor kids / schools buying Model A! At the same time, I was concerned that the difference in profitability would work as a perverse incentive towards limiting the supply of Model A. Of course, the "minority" of hobbyists turned out to be bigger than anyone expected. And, to be honest, nobody is going to convince me that the difference in profitability has played no role in the delay(s) of Model A.
If in 6 months time there is a $0.20 difference between 256M and 512M RAM chips, what is the foundation going to do? Switch to 512M and keep the $25 / $35 price point? Keep it at 256M to discourage people buying a product with lower profit margin? Kill Model A altogether because the 256M version had low demand?
I would seriously consider releasing an "until stock last" model A / 256M at $25. For lots of projects, 256M are more than enough and the advantage of lower power consumption is a compelling feature in itself. That could give the foundation the possibility to properly test a first version of Model A. Then, IMHO, it would make more sense to rise the price of Model A to better reflect its cost and let the market decide what it wants.
Why do I bother with this thoughts? Because I personally think that Model A is technically superior to Model B, and I am concerned about its long term viability. Losing Model A would be a shame.