🚀 go-pugleaf

Johnny Billquist <b...@softjar.se> writes:

> On 2018-07-15 12:17, Johnny Billquist wrote:
>> On 2018-07-15 12:03, Huge wrote:
>>> On 2018-07-15, Johnny Billquist <b...@softjar.se> wrote:
>>>> On 2018-07-15 01:59, Bob Eager wrote:
>>>>> On Sat, 14 Jul 2018 21:22:17 +0200, Johnny Billquist wrote:
>>>>>
>>>>>> Disk directly to buffers in user space in Linux? Sure, it's
>>>>>> called mmap.
>>>>>> Which, by the way, is more or less a direct ripoff from TOPS-20.
>>>>>
>>>>> Which was a ripoff from MULTICS and EMAS.
>>>>
>>>> I'm not sure about that. Multics, from what I read, considered memory
>>>> and files to be the same thing. This is sort of a different idea than
>>>> saying that you can map a file into a specific part of your memory. I
>>>> have no idea how or what EMAS is. :-)
>>>>
>>>> But the comment about the ripoff goes as far as the Unix mmap API being
>>>> pretty much a copy of TOPS-20.
>>>
>>> I find the word "ripoff" problematical. Virtually everything in computer
>>> science is based on something else. Thinking "that's a good idea" and
>>> re-using it in a new and interesting way is hardly "ripping off".
>>
>> If you take the concept *and* API and copy it verbatim, I think
>> calling it a ripoff is not so wrong.
>> Had they just taken the concept, then I'd say inspired by, or
>> something similar. But this one goes a bit further. I remember back
>> when Sun introduced mmap, that I heard (don't remember from where),
>> that they had looked at TOPS-20 when they did the implementation for
>> SunOS.
>
> I should probably modify my language.
> For many, "verbatim" is probably hard to justify here. After all, the
> TOPS-20 PMAP% system call is an API from assembly language, while the
> Unix mmap() system call is an API from C.
>
> Also, PMAP% have some additional features and functions that mmap()
> does not.
> mmap() is maybe more properly described as a subset of PMAP% in one way.
>
> But mmap() is essentially exactly the same as PMAP%.

This may be due to my lack of experience (pretty much all I've ever
programmed for once past toy student projects has been either Unix
derivatives or bare metal) and imagination, but I'm having a hard time
imagining an API for an mmap()-like function that looks substantially
different from mmap().