[Cuis-dev] Questions triggered by #forceChangesToDisk

[Andres Valloud]

Hi, thanks for your comments on the background, I don't know the 
historical context of that VM in detail.

On 9/17/19 02:10, Phil B wrote:
> Just because there was a specification doesn't mean that it was fully or 
> even correctly supported throughout most of the 90's even by those who 
> were supposedly POSIX compliant.

Right, and at the time POSIX was rather new, and those things take a 
while to stabilize.  These days, however, the story is different.

> It's similar to today with HTML standards:  everyone is compliant... 
> kinda, sorta... to varying degrees... with inadvertent and deliberate 
> deviations.

I suspect there's lack of honest effort there... "blah blah ads shine 
brighter in my browser blah blah"...

> And of course everyone knows fflush never fails in the real 
> world (see your famous last words comment ;-)

LOL :).

>     In practice, this means "no MSDN function documentation page will list
>     what errors can occur when calling it", which means "anything goes".
>     This is very much unlike POSIX, i.e. very much unhelpful.
> 
> See my comment about POSIX compliance above ;-)

I meant that I'd rather deal with POSIX than MSDN.

> One thing that can be said in defense of Windows VM development: prior 
> to OS X, Mac VM support was probably just about as distinct.  Mac moved 
> to Unix and Windows stayed put so now it's the odd OS out on this front.

Generally, provided there was a reasonable effort, I can take failures 
and mistakes --- they are unavoidable.  Beyond prioritization, however, 
I see no justification for neglect.

>     Ok, so if that's true, then we're dealing with bit rot.
> 
>     This shows that it is incredibly important to be completely thorough,
>     because it is at that time that a good understanding of the entire
>     problem is in anyone's head.  If you are not thorough today, someone
>     else will have to recreate your state of mind tomorrow.  Overall,
>     everybody goes slower.
> 
> Yep, that's why I'm always complaining about documentation.  What seems 
> obvious today often won't be six months or more from now.  The 
> frustrating thing is that even though we're being thorough, we could 
> still easily be wrong since we're trying to piece together intent with 
> incomplete information.  It's better than guessing but worse than if 
> someone had just written...a ...few...more... words of documentation ;-)

With platform interfaces, I think the best way is to reference and 
follow the relevant platform documentation, because then all you have to 
write is essentially "this source code follows the relevant platform 
manuals, go to https://www.readme.com for details".

Andres.