acroy

I would also say that exceptions provide a clean way to terminate the solver. However, I am not sure that triggering events in the dx/dt callback is a good idea....

Since it just came up in #33: At some point I suggested a simple way to have a [progress callback for Sundials](https://github.com/JuliaLang/Sundials.jl/issues/40#issuecomment-72334904), which might in principle also work for our...

I was actually thinking about adding some examples (and you can find some basic ones in #14 and #10). There is also (at least) one [blog entry](http://phylodynamics.blogspot.co.uk/2013/07/differential-equation-modeling-with.html) on the SIR...

Thanks for the PR! You have accidentally deleted the .travis.yml file, which prevents us from getting Travis feedback. Please put it back.

Is there a fundamental reason why `ode23s` cannot have component-wise relative errors? I think one of the goals of this package is providing a uniform interface for different solvers (if...

Provided type inference works (`reltol` and `abstol` are keywords), separate functions and multiple dispatch allow you to avoid the `if` conditionals, which might have an impact on performance.

@berceanu : AFAIK no one is working on dense output, but it would be a nice feature. It might be that we will replace the current `oderkf` by something evolving...

For DOPRI5 it might make sense to integrate the dense output into the routine, but since we don't have a dedicated `DOPRI5` solver now, I would start with something that...

Yes, I guess you will have to modify oderkf. Maybe one can nevertheless put most of the code into a separate function? And it should ideally work for different RK-triples....

I agree, @tlycken's proposal sounds good. Dense output is mainly (only?) relevant for explicit Runge Kutta methods, so I would say that the general API should only support `: specified`....