* Posts by Mark Morgan Lloyd

10 publicly visible posts • joined 23 Sep 2008

PC pioneer Gary Kildall's unpublished memoir revealed

Mark Morgan Lloyd
Thumb Up

Interesting indeed

That /is/ going to be an interesting read. Kildall did some good work when youger, but more than anything else any insights on how DR became such a pain to deal with in the mid 80s and how they lost their way completely before being bought by Novell will be very welcome.

Parliament takes axe to 2nd EU referendum petition

Mark Morgan Lloyd

Does anybody have a minute-by-minute archive of the .json data, or can say authoritatively where that claim of 39,000 signatures from the Vatican originated?

Linux is so grown up, it's ready for marriage with containers

Mark Morgan Lloyd
Meh

Re: VMs are expensive

> That's partly because both operating systems and CPUs converged on the all-or-nothing

> security model of system space and user space which might have been fine and dandy

> for OS/360 but doesn't really cut it even for client/desktop systems

Seems to me that history is repeating itself: if somebody wanted interactive time on an IBM mainframe he'd generally have ended up with VM, which put a single task into a single virtual machine dedicated to a single user.

Containerisation, which mandates that software be modified to a greater or lesser extent, is definitely one way round the problem. But it seems to me that this leaves a lot of general-purpose software- in particular the sort of scripted hacks that the Internet grew up on- needing a dedicated virtual machine if it's to avail itself of process migration etc. Allowing that VM will always have an overhead in either software or silicon, it's unfortunate that process migration, checkpointing and so on aren't actually part of the kernel.

Dodgy software will bork America's F-35 fighters until at least 2019

Mark Morgan Lloyd
WTF?

Re: @Ledswinger

"...an era when we have oohed and aahed over Windows 95. F35 has been first flown in an era when smartphone meant either Nokia Communicator..."

I think we need to get this into perspective. The Spitfire went from design to first flight in less than two years, and five years after that was in large-scale production and having life-changing results. Now I know that somebody will argue that modern aircraft are vastly more complex than the Spitfire or Harrier, but the Spitfire was designed a mere 35 years after the Wright Brothers first demonstrated powered flight, and quite frankly in the late 30s people knew vastly less about aerodynamics and had vastly less choice of structural materials than they do today.

Add to that that in the late 30s drawings were prepared by hand, any attempt at numerical simulation had to be done by hand, and precision machining was controlled by hand.

So we really should be asking: WTF are we doing wrong? Why is our society unable to build stuff which does the job to the same extent as the Spitfire, Harrier, Apollo program and so on /despite/ our better understanding of the underlying science and our engineering advances? Just about every large-scale aerospace project has the same problems, and most other fields of industry only survive because they're able to recall products and retrofit fixes... and don't anybody get me going on the deficiencies of public-sector IT projects.

-- MarkMLl

Old, not obsolete: IBM takes Linux mainframes back to the future

Mark Morgan Lloyd

Re: Just One Moment...

I think it's not so much things like self-modifying code, as making life comparatively easy for systems software maintainers who regularly have to treat code as data and vice versa. Having mismatched endianness is a serious turn-off for compiler etc. maintainers, and I'd imagine that IBM would far rather not have to devote resources to tweaking other people's codebases- and then having to argue that the tweak was in everybody's interest.

In practical terms, I've seen far fewer problems on big-endian chips running Linux etc. than I have on those architectures with strict alignment requirements, e.g. SPARC and (some) ARM.

p.s. John, like the crypto etc. on your site ;-)

Dear do-gooders, you can't get rid of child labour just by banning it

Mark Morgan Lloyd
Alert

Part of the problem is that in the poorer parts of the world the peasantry don't know enough to realise that giving a child enough food is likely to make them stronger: it stands to reason that a youth less emaciated than his father isn't being worked hard enough.

Debian Project holds Sparc port's hand, switches off life support

Mark Morgan Lloyd
Meh

Re: No surprise there

> SGI kit is waaay too slow, albeit pretty.

SGI gave up on pretty, non-x86 kit a decade or so ago.

> It's a real pity, because in addition to non Intel kit being neat, it's useful to be able to prove

> your software on different endian architectures.

Not just that, but with different alignment requirements etc. Back when compiling stuff on ARM tended to be slow, I was able to report a number of bugs that affected (variants of) the architecture to the Free Pascal Compiler (FPC) developers becuase I could demonstrate and test them quickly on SPARC.

Mark Morgan Lloyd
Meh

Re: Alternatives?

Sun had very little interest in Linux on larger systems, at least until it could be run in a virtualised environment. They wanted to sell Solaris, and made sure that customers bought Solaris by- in part- making sure that it was the only way to get things like environmental monitoring; this is, of course, exactly what IBM does on its bigger iron.

Oracle seem to have little interest in their own Linux targeting SPARC. From their POV, the desktop is x86 and the backend is Solaris.

OpenSolaris (or whatever it's called these days) is available for SPARC in the form of OpenSXCE. Unfortunately the developer of this has strong political opinions, which results in his website being hacked to oblivion on a regular basis. A great pity, since he's done some good work.

As far as "real computers" go, i.e. things that can hold lots of discs, lots of Gb RAM, and lots of CPUs, there's no real alternative to x86 and x86-64. Itanium's dead and was dropped by Debian a year or so ago, MIPS is good for small stuff but it's unclear whether China really does have the balls to push it, and ARM has never really penetrated that part of the industry.

Mark Morgan Lloyd
Meh

The last half-reliable release of Debian on SPARC was Lenny. Superficial examination of Debian's bug tracker shows that there were enough accumulated problems in Squeeze that there should have been no attempt to release anything later.

That is not, of course, to say that pulling everything from the build system is a good idea, since there's already one developer who apparently missed the announcements and is wondering how to get back to a position where he can try to pick up the pieces.

Not everything can be blamed on Oracle. Sun's claim that they'd lost various hardware documentation so couldn't share it with open source developers was pretty laughable, particularly since they were /the/ supplier of computer systems to drawing offices and document management departments for decades.

Data centers embrace The Great Outdoors

Mark Morgan Lloyd

Prior art

In the 1970s a bunch of London hackers with an ICL mainframe occupied what they called the Galdor Center, which was cooled by opening a roller-shutter door. This was fairly widely publicised and probably makes any attempt at patenting the idea difficult.