RE: question about language bindings and elevation

> On 9 Jun 2015, at 23:00, John Jansen <John.Jansen@microsoft.com> wrote:
> > In Windows, when a process such as webdriver spins up a server we have
> an interesting behavior:
> >
> > If the URL is http://localhost:{port} the server spins up fine and webdriver
> can successfully communicate with the browser.
> >
> > If the URL is http://127.0.0.1:{port} the server requires elevation via a
> prompt to "run as admin".
> >
> > Since a large number of WebDriver tests are run on VMs with test
> accounts, I do not want our implementation to require elevation.
> 
> Is this a routing problem in Windows?  127.0.0.1 and localhost are both
> loopback devices.  The point about the 127.0.0.1 or ::1 limitation is simply to
> suggest it’s a good idea that it listens only on locally available devices.

I do believe this is a Windows only issue and has deeply historical implications, so it's not something that will change in the short term. We agree with you regarding locally available devices, but there are real threats.
> 
> > However, it is not clear to me what priority I should give the task to enable
> this scenario. The only language bindings I have seen that use the IP address
> rather than the friendly name are the Python ones.
> 
> Actually the reason Python uses the numeric IP address is IIRC because a
> user reported problems using localhost on some systems, although if my
> understanding of loopbacks isn’t misguided “localhost” and “127.0.0.1”
> should be equivalent, i.e. they are loopbacks.

OK, if there will be new issues with using localhost instead of IP, then we won't bother going down this route.
> 
> > 2. What does this group think about Microsoft Edge shipping an initial
> implementation that requires elevation when the Python language bindings
> are used?
> 
> My gut reaction is that this is a problem elsewhere and that shipping would
> not be a problem.

Thanks for the feedback.
>

Received on Wednesday, 10 June 2015 21:58:10 UTC