CTFE ^^ (pow)

Paolo Invernizzi paolo.invernizzi at gmail.com
Mon Mar 19 08:38:17 UTC 2018


On Monday, 19 March 2018 at 05:27:20 UTC, Norm wrote:
> On Monday, 19 March 2018 at 04:15:26 UTC, rikki cattermole 
> wrote:
>> On 19/03/2018 5:05 PM, Norm wrote:
>>> On Monday, 19 March 2018 at 03:53:07 UTC, rikki cattermole 
>>> wrote:
>>>> On 19/03/2018 4:43 PM, Norm wrote:
>>>>> [...]
>>>>
>>>> You just said the magic word, medical.
>>>>
>>>> D was never an appropriate fit here.
>>>>
>>>> dmd's backend has been for thirty years (or so) been up to 
>>>> recently licensed so that you may not use it for this 
>>>> purpose. Nothing has changed here.
>>> 
>>> I have no idea what you're talking about now.
>>> 
>>> What has the backend license got to do with medical?
>>
>> The code generation capabilities of dmd has not been certified 
>> for medical usage.
>>
>> In essence, if it generated bad code, kills somebody, your the 
>> one at fault, even if the source is fine. You would end up 
>> begging to settle out of court.
>>
>> It is my understanding that medical software manufacturers pay 
>> for their compilers already certified. So that suggests to me 
>> that you're not exactly life threatening but I would still 
>> caution you away from D even if that bit is just my own 
>> opinion.
>
> No, compilers do not need to be certified for class B or class 
> C software. These are the two highest safety classes for 
> medical SW. Beyond class C SW is not allowed, e.g. safety 
> critical interlocks such as the big red button to shut off a 
> radiation dose or stop a robotic system.
>
> Compilers are are treated as SOUP (Software of Unknown 
> Provenance), i.e. a black box. Risk analysis leads to risk 
> control measures that in turn ensure people don't die and this 
> is done at the system and component level, not the codegen 
> level. Verification is performed to ensure the system 
> implements the requirements correctly, and subsequently the 
> risk control measures. Not all requirements are risk control 
> measures, but all requirements must be verified as correct.
>
> Cheers,
> Norm

I was the CTO and partner of a company using D in medical devices 
since more than ten years ago... as Norm wrote, medical software 
is a strange beast...

Anyway, as someone else wrote, when I leaved the company, two 
years ago, the new CTO and my former colleague, decided not to 
invest anymore in D. After ten years of use.

Said that, I'm pretty happy about what's happening in D Land in 
the last 3/4 months, but clearly there's a lot to be done.

/Paolo









More information about the Digitalmars-d mailing list