Re:Microsoft's “Love” of Linux
by asveikau on Thursday November 21, 2019 @03:28PM
(#21599191)
> putting com interfaces into kernel32.dll
I think I read this claim from you here before. What's up with that?
These days, kernel32 (split off into kernelbase and others in the Vista and Win7 days) mostly has small wrappers to ntdll.
A COM interface is really just a C++ VTable. So the "overhead" of that is basically minimum 3 function pointers and some indirect calls. What is the problem with having a COM object passed to a kernel32 function? Would you have the same objection if it were a (barely any different) function pointer plus an extra pointer for context?
The complaint about manifests maybe made me think this was about WinSXS or something. Which I would agree is engineered kind of poorly, but I doubt it's the worst thing in Windows.
[Disclaimer. I worked as an IC SDE at MS from 2008-2011.]
I think I read this claim from you here before. What's up with that?
These days, kernel32 (split off into kernelbase and others in the Vista and Win7 days) mostly has small wrappers to ntdll.
A COM interface is really just a C++ VTable. So the "overhead" of that is basically minimum 3 function pointers and some indirect calls. What is the problem with having a COM object passed to a kernel32 function? Would you have the same objection if it were a (barely any different) function pointer plus an extra pointer for context?
The complaint about manifests maybe made me think this was about WinSXS or something. Which I would agree is engineered kind of poorly, but I doubt it's the worst thing in Windows.
[Disclaimer. I worked as an IC SDE at MS from 2008-2011.]