The code would be just as well perfectly convenient for the user if
uploaded to Akku.SCM or snow-fort.
They are as portable as portability can be achieved in Scheme, and do
not require such a rigourous scrutiny as srfis do.

On Fri, 17 Jul 2020 at 10:26, Linas Vepstas <xxxxxx@gmail.com> wrote:
>
> Hi Vladimir,
>
> That's maybe a reasonable comment for sphere distributions, which are relatively easy to implement for those people who can read math, but removing something like  the gaussian distribution is too deep a cut.  The point is that you could roll-your-own sphere in an hour or two.  But trying to create gaussian or poisson or zipf from scratch turns out to be a huge amount of work, a huge distraction, if you just want to "get some work done".
>
> I look at the srfis from the point of view "does it make things easier for an ordinary user who just wants to *use* scheme for ordinary things?" And for the ordinary user, who just wants to use things without thinking hard, without working a lot, the srfi's are ideal. They offer tested, working code. Admittedly, you still have to read the documentation, but that is a price worth paying.
>
> To create a "flow-blown statistical CAS" would take what -- 10 or 20 people several years to do? We've got exactly two people working on a CAS system, they've been at it for 2-3 years now, they've got a small handful of basics done, but its absurd to suggest that they should also become experts on statistics. They're not educated in that kind of stuff.
>
> --linas
>
> On Thu, Jul 16, 2020 at 9:11 PM Vladimir Nikishkin <xxxxxx@gmail.com> wrote:
>>
>> Hello, friends
>>
>> I'm not sure I am the best person to comment on this, as I'm not a
>> scheme implementor (well, I am, but I don't intend to develop
>> schemetran further).
>>
>> To be honest, I do not see a valid reason for such an extension to
>> Scheme as a language. Clearly, "some" sort of random number generation
>> is crucial for a decent programming language.
>>
>> However, there is a difference between "basic needs of a programming
>> language" and a full-blown statistical CAS, and the discussion that is
>> happening in the mailing list looks way more like one fit for the
>> discussion for a statistical CAS than a language extension.
>>
>> (FWIW, If implementing a CAS, I wouldn't like any pre-built generators
>> at all, I would prefer a good data structure for representing
>> distributions, a mean of combinations of those, and a completely
>> independent set of sampling algorithms, accepting those distributions
>> as parameters in order to produce generators.)
>>
>> I understand that this srfi is essentially a documentation of an
>> existing practice, and is not expected to be perfect, but if a
>> portable r7rs library already exists, is implementable using already
>> portable extensions, why should it be an srfi?
>>
>> So, in general, even though random numbers could be a good use case
>> for generator application, I do not think that this srfi is an example
>> of good design, and if its aim is to be an illustration of srfi-27 and
>> srfi-154, I would rather remove all non-uniform distributions
>> altogether, as this would factor out a lot of questions regarding
>> statistical rigour. (Which is itself a huge, and still incomplete
>> research area.)
>>
>> Sorry for being so skeptical.
>>
>> --
>> Yours sincerely, Vladimir Nikishkin
>
>
>
> --
> Verbogeny is one of the pleasurettes of a creatific thinkerizer.
>         --Peter da Silva
>

--
Yours sincerely, Vladimir Nikishkin