Re: [Boost.Asio] service_name/port inconsistency

On Fri, 29 Aug 2008, Vjekoslav Brajkovic wrote:
> Hi.
>
> When designing server-client model, one has to use
> boost::asio::ip::tcp::endpoint and boost::asio::ip::tcp::resolver::query
> respectively. What I am confused about is that query accepts
> std::string& service as a parameter whereas endpoint only accepts
> unsigned short.
>
> When writing code, it is really easy to get confused which constructor
> takes string and which short. I would appreciate is somebody could
> tell me if there is a way to create a server side component using
> integer as a service.

Sorry, not the server, but the client. Also, let extend my question. It
seems if one wants to resolve a FQDN, first ip::basic_resolver<T> has to
be created. Afterwards, basic_resolver.resolve() function should be called 
which accepts either endpoint_type or query as a parameter. So what 
I was wondering is:

a) What is the point passing endpoint_type, when we know for sure that
endpoint will only accept address which *cannot* be a FQDN. What do we
achieve by doing this?

b) To be more precise: How come basic_resolver_query does not take
unsigned short as a parameter for service/port?

I apologize if these are somewhat broad questions, however I really would
like to know the rational behind some of the design decisions.