expanded a bit more
In the old Sullivan reference the condition for A to be formal is that there is a quasi-iso .
In newer references, such as the survey by Kathry Hess, it says instead that there is a span of quasi-isos between and . Since in the model structure on dg-algebras all objects are fibrant, I take it that this is equivalent to saying that and are isomorphic in the homotopy category. This is the way I state the definition now. But check if I am mixed up.
]]>started formal dg-algebra
]]>