[Ns-developers] NS3 Time objects

[Tom Henderson]

Mathieu Lacage wrote:

> 
> I pushed the final version with doxygen tagging, documentation and
> regression tests in http://code.nsnam.org/mathieu/ns-3
> 
> I would like to go forward with this set of features so, baring further
> comments, I would like to push this implementation in the main repo next
> week. I think we can get the 128bit integer version in later: it should
> have zero impact on user code.
> 

Mathieu, although I am not sure that some of these use cases will crop 
up in practice (regarding timesquared and other higher-ordered 
operations), they do not seem to complicate the user API for the basic 
user, so I agree with your proposal and for pushing it in.

I think it is very similar to George's proposal from 10/30 except for:

- mixing data types in operators is avoided, which I think seems to be 
the safer option to avoid commutativity and associativity problems

- you supported a use case (multiplication and division of time objects) 
in which George was not interested; I don't care strongly either way

- as for string conversions, I don't have a strong opinion of whether 
these should be member functions or not; I would probably lean towards 
non-member functions for encapsulation reasons

One thing that I think you should not do is punish the user so severely 
for trying to convert Time to a double ("ApproximateToSeconds").  I 
would suggest "GetSeconds" and just document the function with a doxygen 
comment that object to float type conversions may be subject to certain 
portability issues.

Thanks for doing all of the doxygen up front too,
Tom