info on partitioning in survival trees

Question

I'm hoping to use survival trees for my dataset, where so far I mainly used proportional hazards models. I'm interested in obtaining quantiles of the survival function. I'm struggling a bit to understand how survival trees partition the data and what the actual output of the model is.

in rpart's manual i couldn't find much than this:

My guess is that rpart tries to do is either fit Kaplan Meier to make the splits using log-rank test, or fit a proportional parametric hazard with exponential distribution, and has a different way of splitting it?

Can I ask for help in unpacking this?

1. what are the poisson methods that rpart.exp uses?
2. what are "shrink" and "method" in rpart.exp's params?
3. Quantiles of the survival function- I believe this requires modelling of the distribution of survival times. Is this happening here?

Or where can I get more information?

thanks!