Long uptimes.

Levi Pearson levipearson at gmail.com
Wed Nov 18 10:55:16 MST 2009


Michael Torrie <torriem at gmail.com> writes:

> Levi Pearson wrote:
>> So, bascially what you're saying is that anyone who chooses to implement
>> a different set of features than you'd like is inherently lazy?  I'm
>> sorry the world hasn't arranged itself to suit your needs. :)
>
> In fact I am, although I'm not referring to my personal computing needs.
>  The total economic cost to the world economy caused by requiring
> reboots is probably many times that of the cost of fixing the problem
> right originally.  But it's a matter of who bears the cost.

The total economic cost to the world economy caused by requiring reboots
may indeed be significant in total dollars, but I'm pretty sure it's
lost in the noise in a complete economic analysis of costs.  You'll have
to show me some real numbers, not silly Steve Jobs style thought
experiments, to make me believe otherwise.

When it comes down to it, the people who bear the costs don't consider
them great enough to demand that particular feature, or else they would
pay up front to have it implemented.

>> OK, considering we're all pretty much running some form of a unix or
>> windows system, none of which have a kernel architecture newer than
>> sometime in the 90s, and some of which go back to the 70s... wouldn't
>> you say that their development is focused on software maintenance?
>
> Well, read the article.  He makes it pretty clear that at one time
> software maintenance was the focus, but not anymore.

I read the article.  He made it clear that at one time, people did
software maintenance differently than they do now, in general.  But his
primary example is MULTICS, which existed in an entirely different
computing 'ecosystem' than exists now.  And he also listed several
modern elegant solutions to maintenance.

Clearly Linux, OS X, BSD, Windows kernel, etc. are being maintained.
They are evolving to meet new needs, new hardware, etc.  Just because
they can't be upgraded seamlessly doesn't mean there's no focus on
maintenance, it just means they haven't determined that preventing the
need for reboots is a necessary goal for them.

For desktop systems, rebooting is not a significant cost, no matter how
much it irks you.  For most servers, it's not a big deal either, and if
a single node's downtime would be expensive, you should have a redundant
backup anyway and then you don't have any real downtime if you stagger
the reboots.  The fact is, the software and hardware we have now have
evolved pretty well to meet the actual needs of the people who pay for
them.  Are they optimal?  No, they kind of suck in a lot of ways.  But
that's not because people are lazy, it's because they're not in it for
elegance, they're in it to use these things to make money.

>> Maybe they're just driven by some metric other than ridiculously long
>> uptimes.  It would make sense, since to do so would be... ridiculous.
>
> Yes I was afraid I'd be misconstrued as saying that ridiculously long
> uptimes (as in the kind reported by netcraft) are the goal.

The goal is apparently to meet your standards.  That's fine, but you'll
only be disappointed.

                --Levi



More information about the PLUG mailing list