On Wed, Oct 3, 2018 at 11:08 AM Michael Osipov <[email protected]> wrote:

> Am 2018-10-03 um 18:38 schrieb Gary Gregory:
> > On Wed, Oct 3, 2018 at 10:18 AM Michael Osipov <[email protected]>
> wrote:
> >
> >> Am 2018-10-03 um 16:55 schrieb Gary Gregory:
> >>> On Wed, Oct 3, 2018 at 7:17 AM Oleg Kalnichevski <[email protected]>
> >> wrote:
> >>>
> >>>> On Tue, 2018-10-02 at 15:34 -0600, Gary Gregory wrote:
> >>>>> On Tue, Oct 2, 2018 at 2:09 PM Oleg Kalnichevski <[email protected]>
> >>>>> wrote:
> >>>>>
> >>>>>> On October 2, 2018 9:57:40 PM GMT+02:00, Michael Osipov <
> >>>>>> [email protected]> wrote:
> >>>>>>> Am 2018-10-01 um 16:33 schrieb Gary Gregory:
> >>>>>>>> Also, since we are here, in the same place: "Header
> >>>>>>>> getSingleHeader(String)", could be be "getHeader(String)";
> >>>>>>>> since the
> >>>>>>>> "Single" is implied by the singular "Header" in the API name
> >>>>>>>> _and_ by
> >>>>>>>
> >>>>>>> the
> >>>>>>>> return type.
> >>>>>>>>
> >>>>>>>> Thoughts on that one?
> >>>>>>>
> >>>>>>> Makes sense to me.
> >>>>>>>
> >>>>>>
> >>>>>> That will break a lot of people's code. Please at the very least
> >>>>>> deprecate
> >>>>>> all methods.
> >>>>>>
> >>>>>
> >>>>> IMO, deprecating code from one beta to another is over the top and
> >>>>> when do
> >>>>> you remove the methods, in another beta?
> >>>>>
> >>>>
> >>>> I imagine such methods could be removed at any point before the first
> >>>> GA release.
> >>>>
> >>>> I do not understand why making an upgrade to the next BETA simpler for
> >>>> every single HC 5.0 user out there is not worth one extra commit.
> >>>>
> >>>
> >>> Good idea. In git master:
> >>>
> >>> * Deprecate and rename
> >>> org.apache.hc.core5.http.EndpointDetails.getSocketTimeout() to
> >>> getSocketTimeoutMillis().
> >>
> >> I am not a huge fan of having the value scaling as part of the method
> >> name. An overload as getSocketTimeout(TimeUnit) would not work anymore.
> >>
> >
> > Not having the scale is a recipe for disaster -- I'm being dramatic :-)
>
> That is just bad documentation.
>
> > We already have scale in other APIs.
> >
> > Nothing prevents you from adding getSocketTimeout(TimeUnit), so what if
> > it's not an overload? What matters is that it would be an easy API to
> > understand IMO.
>
> I do simply mean that getSocketTimeout() and getSocketTimeout(TimeUnit)
> read way better thatn getSocketTimeoutMillis() and
> getSocketTimeout(TimeUnit). It just looks consistent.
>

Ah, well, that's a matter of opinion ;-)

Gary

Reply via email to