[ProgClub list] So, why aren't we using Unix as god intended?

Asher Glynn asher.glynn at gmail.com
Thu Oct 27 00:55:06 AEDT 2011


On 26 October 2011 14:34, John Elliot <jj5 at progclub.org> wrote:

> On 27/10/2011 12:05 AM, Asher Glynn wrote:
>
>> Its just an interesting dilemma - on the one hand Unix is crafted and
>> guided
>> by many many design decisions that actually turned out to be bollocks in
>> the
>> long term, so we're living with a design that no longer fits our usage
>> pattern.
>>
>
> I think I agree with you. I'll look forward to seeing your article.
>
Focusing on short(ish) posts

http://asher-glynn.blogspot.com/2011/10/unix-on-hypervisors-what-have-you-done.html



> I've been thinking along similar lines about Blackbrick's (planned :P)
> Homepage Server. That will support multiple concurrent users, but the
> security will be managed at the application layer, and not by deferring in
> any way to the (Linux) platform. From the operating system's perspective
> that will be the "single user running single daemon process" that you
> alluded to earlier, essentially www-data running Apache. Actually, in
> practice, I expect it will be more than that, but not by much.
>
> I'm happy with *nix because I can use its intended idioms when it suits me,
> and ignore them if it doesn't. Sometimes I use my *nix systems in all the
> ways you enumerated as intentions, and sometimes I don't. I guess it's a
> matter of trying to pick the right tool for the job. I feel like the Unix
> approach left me with options.
>

I agree - Unix is the correct default option, but why don't we have well
known alternatives?


>
> A question that springs to mind, and I'm surprised I haven't got an answer
> for this off the top of my head, but what are the design decisions taken by
> Windows that differ from *nix? Is it simply that Windows is more graphically
> oriented, or does it go further than that? If you think *nix is out-dated,
> what is your opinion on Windows (or OSX, to the extent it's not just BSD
> Unix)?


Windows was based around the experiences of the VMS guys that came from
Digital - very clever guys. MS benefited from them in the same way as a
whole lot of pissed off Digital CPU designers benefited AMD for a bit,
bringing specific skills in. Hangovers in Windows from OS/2 are based around
some of IBM's mainframe experience and DOS being a variant on CP/M which
actually hails in part back to CP/CMS in the mainframe. The "official"
language for CP/M was PL/M which was a variant of PL/I, definitely a
mainframe programming language. Windows NT is interesting because its
actually a much more modern design than Linux, but I think it suffers
because it is too much of a hodge podge of design concepts.

OS X is interesting - they made a very good decision to base on something
innocuous like BSD rather than what they were originally using at NeXT but
you can see the years the poor guys at NeXT really thinking hard about
things burning through into OS X. If anyone gets hung up about Objective C
being used for it - go find a mirror and try and tell yourself that C++ in
any way is a well thought through language and watch the mirror crack itself
up laughing.

Picking an existing starting point for a kernel is completely legit, you
just have to live with all the other bits of cruft you inherit with it.

To squeeze performance from hypervisors we're increasingly using specific,
dedicated network drivers and disk drivers that are just thin shims on
hypervisor services - not a hundred miles from either microkernel OSes
ironically.

The real alternative would be something like the Hurd - but that frankly
requires a few beers to take even remotely seriously.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.progclub.org/pipermail/list/attachments/20111026/d6d120dd/attachment.html>


More information about the list mailing list