Re: [tlug] Unix's 40th Birthday

["Stephen J. Turnbull" <stephen@example.com>]

Curt Sampson writes:

 > Incidently, that "big iron" is not so big in terms of processing
 > power, it's just distributed and used in a very different way.)

That's always been true of big iron.  Big iron has always been about
I/O, not processor power or even memory.  (At least for commercially
demanded models.)

 > True enough, though you might be suprised at the number of firewalls,
 > VPN servers, and so on that run on Cisco-like gear.

But what OS does that run?  I wouldn't be surprised if it's a Unix
derivative, what is it?  Those do need to be multi-tasking at the user
process level.  They can't be more or less stateless packet handlers.

Even your dedicated Cisco router ... surely it's not running a DOS or
NT derivative, the Internet couldn't function for 5 seconds under load
if it were!  I suppose they could have developed one from the ground
up, but I'd think it would be a lot cheaper to just use NetBSD and
replace the scheduler with something realtime.

 > For home use, however, I'd expect that a lot of VPN connectivity,
 > for example, is done by a Windows box.

Could be.  However, everybody I know who does that has an ISP-supplied
broadband router/gateway, and possibly a wireless basestation, and
does the VPN on only one box that connects through the router.  I
don't really consider that box "part of the Internet" in the sense
we're talking about here; it doesn't do any routing.

My statement was hyperbole, of course.  However, I think it's
important to 

 > However, I think that MS, even lacking the better leadership that
 > it had in the 90s, has a reasonable chance of coming to their
 > senses at some point, and they do have a huge market of people who
 > will buy the product regardless to support them while they come to
 > their senses.

That's why I mention "disruptive innovation".  "Disruptive" doesn't
mean "radical".  An innovation is disruptive when (1) it finally
crosses the threshold of being an interesting alternative to the
"mainstream" configuration for customers, but (2) is too cheap and
competitive to sustain profit margins that can sustain earnings growth
in the large incumbent market leaders.

 > It's also going to be interesting to see if the competition can avoid
 > the downhill slide that's started over the last couple of years. Every
 > time I upgrade Ubuntu it gets a bit more byzantine,

This is going to continue, I'm afraid.  If Linus were interested in
system security ... but unfortunately what we've got is Theo, who I
supose is very competent but gathers enemies the way the north side of
a tree gathers moss.  And Theo is on the 'wrong' platform to excite a
lot of interest from GNUbies....

I don't know WTF the big distros are thinking.  They really ought to
put some resources into straightening out the PAM mess, for one
thing.  That web site is horrifying, both for the quality of the
documentation itself, and for the design whose brokenness can't be
hidden by the docs.  It wouldn't cost that much.

 > And if the Ubuntu and Debian maintainers are heading towards a
 > Windows level of security, what's going to happen as more and more
 > vendors start adapting Linux to their custom boxes?

That's hard to say.  Most of my students don't know how to use a PC
(ie, they treat it as a large-screen keitai with a full keyboard and
mouse).  I think that in countries without an existing PC culture
(including Japan, the tigers, and the BRICs, as well as the countries
still mired in the Third World) PCs probably will *never* achieve the
penetration they have in the US and Europe.  The keitai is already a
net appliance, and most people will demand that from their
workstations as well.

It's not obvious to me that Windows has an advantage in that kind of
world other than Microsoft's huge R&D budget.  I mean, consider what
happens when no-user-servicable-parts Google Chrome migrates from the
Android to the desktop.  Don't even *think* about the possibility that
Google *isn't* thinking about it.

 > Well, that definition changes rapidly, of course. For mid-range
 > developer boxes, 3GB has been the minimum memory size (you need 3 DIMMs
 > for all but the newest i7 motherboards) for a year now, and these days
 > we wouldn't bother to put in less than 6GB.

Sure, but is that really a market that Microsoft can afford to try to
squeeze profit from?