Feedback Thread: DIP 1042--ProtoObject--Community Review Round 1
Ola Fosheim Grøstad
ola.fosheim.grostad at gmail.com
Sat Jan 15 22:57:36 UTC 2022
The DIP is jumping into details without acknowledging that the
class concept is a high level construct related to OO
methodology. The design should be rooted in a rationale based on
specific OO modelling principles (maybe read conference
proceeding for ideas) and usability.
Since it is a high level concept you essentially should strive to
make it easy on the programmer and put the burden of optimization
on the compiler. This must be discussed. For instance, an
implementation can remove the mutex if it isnt used. It can even
do so in separate compilation by putting the mutex at a negative
offset. If you need more details, feel free to ask.
All other OO languages in the Simula family use «object» as the
root (if they provide one). Deviating from this norm will cause
confusion and create friction and annoyance among people who have
an OO background. The motivation for deviating from this is not
It is desirable to have one root, yet, you propose 3 roots? This
needs more motivation. D needs to be made simpler, not more
I am not getting into the specifics of interfaces, but as
implemented they are to be avoided as they cost 8 bytes per
instanced object per interface. So it is better to adorn the root
object with abstract virtuals which are free. If you want to
force interfaces on developers then you need to provide an
interface mechanism that is usable (the current one is too
The list of popular languages at the end does not seem motivated.
Why only these commercial languages and why listing them? By not
referencing more academic OO resources the reader will get the
impression that you did not do a proper survey and that the DIP
as written is more of an early sketch than a final design.
More information about the Digitalmars-d