On 14 October 2011 10:15, Troy Straszheim <straszheim@willowgarage.com>wrote:

> On Thu, Oct 13, 2011 at 5:57 PM, Daniel Stonier <d.stonier@gmail.com>
> wrote:
> > Having TCP_NODELAY as a non-default option at ros::init might be
> convenient.
>
> Globals...   My subscribers' behavior can change if somebody else
> calls ros::init on my behalf...
>

Oh, like a library written with certain specs in mind, then a user running
that library but setting a different ros::init global option and thus
affecting behaviour of the library in a non-desired way?

Got it.

Daniel.