Re: [Boost-users] Design for extensibility

From: Tim St. Clair (timothysc_at_[hidden])
Date: 2007-09-07 10:32:10

Eric -

     The thread class is a good example, by having the handles private
instead of protected there is no way to extend the class.  If it were
#defined wrapped for extensibility that would be groovy too.

     I'm looking to extend an interface which presents a pthread-esk
interface, while supporting multiple OS's (VxWorks, Linux, Windows,
Other Embedded OS's ).  So our application code can agnostic.

-Tim

p.s. - BOOST_FOREACH is a jem!

On 9/6/07, Eric Niebler <eric_at_[hidden]> wrote:
> Tim St. Clair wrote:
> > Folks -
> >
> >       I have found that there are several classes in boost, namely the
> > OS constructs (thread, mutex, etc.) do not provide me w/some of the
> > features I'm looking for (and that is ok with me).    As an embedded
> > developer I deal with all kinds of operating systems which are not in
> > the main stream, however C++ is very common in the embedded world, and
> > is growing.
> >
> >       With that note, I would like to assert the following:
> >       - It would be "uber-useful" to design OS constructs, or
> > libraries for that matter, with extensibility in mind.  By providing a
> > couple of guidelines, developers could then port to any OS under the
> > sun, or add the features they desire.
> >
> >       Instead, I'm left with writing yet another wrapper (ohh joy).
>
>
> Nobody can really do anything with this feedback until you tell us
> specifically which classes you experienced problems with, what the
> problems were exactly, and what a better design would be, in your opinion.
>
>
> --
> Eric Niebler
> Boost Consulting
> www.boost-consulting.com
>
> The Astoria Seminar ==> http://www.astoriaseminar.com
> _______________________________________________
> Boost-users mailing list
> Boost-users_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-users
>

-- 
Regards,
Timothy St. Clair
[timothysc_at_[hidden]]